Elderly NSC emergency treatment risk layering method, device and medium

Abstract

The invention relates to the field of clinical diagnostics, and discloses an elderly NSC emergency risk layering method and device and a medium, and the method comprises the steps: S1, obtaining multi-dimensional evaluation data of an elderly NSC patient; s2, adopting a preset risk mapping rule to convert the risk values into single risk values of a unified scale; s3, dividing the single risk value into a plurality of risk sub-models, and calculating a dimension risk score of each risk sub-model; s4, based on a preset fusion strategy, integrating the risk sub-model and the single risk value interaction effect, and calculating to obtain a total risk score of the patient; and S5, determining the risk level of the patient according to the total risk score, and outputting a clinical diagnosis and treatment suggestion. According to the method, multiple biomarkers and key clinical parameters are integrated, a multi-dimensional combined risk assessment model is constructed, and the potential pathological state of a patient can be reflected more comprehensively.

Description

Technical Field

[0001] This invention relates to the field of clinical diagnostics, specifically to a method, device, and medium for risk stratification in emergency care for elderly patients with NSC. Background Technology

[0002] The elderly patient population presenting with non-specific complaints (NSCs) in the emergency department is well-known but poorly defined (simple complaints of feeling unwell, dizziness, etc.). This group is often difficult to diagnose and poses a significant challenge to emergency physicians. Studies have shown that up to 20% of elderly emergency room patients present with non-specific complaints, and approximately 50% of these eventually develop acute medical problems. Furthermore, elderly patients are inherently a high-risk group for adverse outcomes such as functional decline, dependence, and death.

[0003] In the emergency department setting, multiple biomarkers hold potential for risk assessment. suPAR (a broad-spectrum inflammatory activity marker), a non-specific biomarker reflecting systemic inflammatory status and disease prognosis, has been shown to be an important indicator for risk stratification in emergency patients. Studies have shown that introducing suPAR can improve the performance in predicting short-term mortality in emergency patients. Furthermore, MR-proADM (mid-terminal pro-adrenergic glycoside) and D-dimer have also shown excellent performance in predicting mortality risk in elderly emergency patients. In univariate logistic regression analysis, the odds ratio (OR) for 30-day mortality was as high as 11.12 (approximately 7.65 after multivariate adjustment) with elevated MR-proADM levels, and the corresponding area under the curve (AUC) was approximately 0.86; while the AUC for D-dimer at the standard cutoff value also exceeded 0.85. These results suggest that multi-marker models can significantly improve predictive accuracy.

[0004] In summary, there is still a lack of comprehensive solutions for risk assessment of elderly NSC patients, and the selection of biomarkers and algorithm design of existing methods need to be optimized and innovated. Summary of the Invention

[0005] In view of this, the present invention provides a method, device and medium for risk stratification in elderly patients with NSC emergency care to solve the above problems.

[0006] To address the above technical problems, this invention provides a method for risk stratification in emergency care for elderly patients with NSC, comprising: S1: Obtain multidimensional assessment data of elderly NSC patients and standardize the multidimensional assessment data; S2: For each standardized assessment data point, a pre-defined risk mapping rule is used to convert it into a single risk value on a uniform scale; the risk mapping rule is set according to the pathological significance of each assessment data point and the physiological characteristics of elderly NSC patients; S3: Divide a single risk value into multiple risk sub-models and calculate the dimensional risk score for each risk sub-model; S4: Based on the preset fusion strategy, the dimensional risk scores of multiple risk sub-models and the interaction effects between individual risk values ​​are integrated to calculate the patient's total risk score. S5: Determine the patient's risk level based on the total risk score, and output corresponding clinical treatment pathway recommendations based on the risk level.

[0007] As an optional approach, multidimensional assessment data should include at least basic clinical information, laboratory biomarker indicators, and vital sign indicators; Laboratory biomarkers include at least one of the following: soluble urokinase-type plasminogen activator receptor, adrenal medullaris precursor mid-terminus, D-dimer, vasopressin C-terminal precursor, and peroxidase-4.

[0008] As an optional approach, standardization is achieved using the Z-score algorithm, which eliminates differences in measurement units and dimensions among different evaluation data. This includes: Based on the mean and standard deviation of each assessment data in the historical training data, the original values ​​of the current patient's assessment data are transformed and standardized.

[0009] As an optional approach, the preset risk mapping rules include nonlinear mapping functions, and specific function forms are adopted for the pathological mechanisms of different assessment data; Among them, a piecewise quadratic function is used for indicators reflecting broad-spectrum inflammatory activity, a sigmoid function is used for indicators reflecting organ perfusion status, a power function is used for indicators reflecting the degree of functional decline, and a nonlinear function based on the normal range is used for indicators reflecting heart rate.

[0010] As an alternative, the risk mapping rule also includes an age-based dynamic correction threshold, which is used to distinguish between normal and abnormal risk ranges for D-dimer.

[0011] As an optional approach, the output range of a single risk value with a uniform scale is set with a preset interval. The values ​​within the preset interval are used to realize the probabilistic expression of the risk of a single assessment data, and the key parameters of each risk mapping rule are automatically optimized based on the statistical distribution of historical training data.

[0012] As an alternative approach, multiple risk sub-models include sub-models based on five pathological mechanism dimensions: inflammation-immunity sub-model, coagulation-circulation sub-model, organ function sub-model, clinical perception sub-model, and basic risk regulation sub-model. Among them, the inflammation-immunity sub-model is used to integrate individual risk values ​​related to inflammation and immunity; the coagulation-circulation sub-model is used to integrate individual risk values ​​related to coagulation and cardiovascular function; the organ function sub-model is used to integrate individual risk values ​​related to organ perfusion and respiratory function; the clinical perception sub-model is used to integrate individual risk values ​​related to the degree of functional decline and the doctor's initial impression; and the basic risk regulation sub-model is constructed based on the patient's age.

[0013] As an optional approach, the dimensional risk score of each risk sub-model is a weighted linear combination of all individual risk values ​​under that sub-model; The weight of each individual risk value is calculated based on historical training data, and the sum of the weights of all individual risk values ​​within each sub-model is 1.

[0014] As an optional approach, the preset fusion strategies include: The first fusion layer is set as a global weighted sub-model, and the dimensional risk scores of the five risk sub-models are linearly aggregated through global weights; The second fusion layer is set as an interaction item penalty, which corrects the individual risk value pairs with synergistic deterioration effect through interaction weights. Individual risk value pairs with synergistic deterioration effect include one or more of the following: suPAR and CRP, MR-proADM and respiratory rate, blood oxygen saturation and systolic blood pressure.

[0015] As an optional approach, the risk level includes three levels: low risk, medium risk, and high risk. Each level has a different total risk score range. The threshold of the score range is dynamically updated based on the hospital's local historical data. The update method is to determine the optimal threshold by calculating the Youden index through ROC curve analysis every quarter.

[0016] On the other hand, the present invention also provides an electronic device, comprising: a memory for storing a computer program; and a processor for implementing a method for stratifying emergency risk in elderly patients with NSC when executing the computer program.

[0017] On the other hand, the present invention also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of a method for risk stratification in emergency care for elderly patients with NSC.

[0018] The beneficial effects of this invention are as follows: This invention integrates multiple biomarkers and key clinical parameters to construct a multi-dimensional joint risk assessment model, which can more comprehensively reflect the patient's potential pathological state. Compared with traditional methods that rely on a single indicator or scoring system, this model shows stronger discriminative ability in identifying the potential risk of severe illness in elderly patients with nonspecific chief complaints, especially in the early stages when clinical manifestations are atypical, demonstrating higher identification value. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the elderly NSC emergency risk stratification method of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been presented in the various embodiments of the present invention to enable the reader to better understand the present invention. However, the technical solutions claimed in the present invention can be implemented even without these technical details and various changes and modifications based on the following embodiments.

[0021] like Figure 1 As shown, this specific implementation method details the technical implementation process of an emergency risk stratification method and system for elderly NSC patients based on multi-parameter biomarker fusion. It is applicable to emergency patients aged 65 and above with nonspecific complaints, often presenting with symptoms such as malaise and dizziness without a clear target. This embodiment, through multi-dimensional data collection, standardized processing, risk mapping, sub-model construction, overall risk fusion, and linkage with treatment pathways, can achieve accurate prediction and stratified management of patients' short-term severe adverse outcomes. This embodiment is implemented as follows: S1: Obtain multidimensional assessment data of elderly NSC patients and standardize the multidimensional assessment data; S2: For each standardized assessment data point, a pre-defined risk mapping rule is used to convert it into a single risk value on a uniform scale; the risk mapping rule is set according to the pathological significance of each assessment data point and the physiological characteristics of elderly NSC patients; S3: Divide a single risk value into multiple risk sub-models and calculate the dimensional risk score for each risk sub-model; S4: Based on the preset fusion strategy, the dimensional risk scores of multiple risk sub-models and the interaction effects between individual risk values ​​are integrated to calculate the patient's total risk score. S5: Determine the patient's risk level based on the total risk score, and output corresponding clinical treatment pathway recommendations based on the risk level.

[0022] In S1 above, multi-dimensional assessment data needs to be automatically collected from the hospital information system, laboratory information system, and electronic medical records to ensure data integrity and consistency. The core data collected is divided into three categories. Basic clinical information indicators include age, ADL score, and physician's first impression score. Age must be 65 years or older, serving as the core for risk baseline adjustment. The ADL score is used to assess the degree of functional decline, ranging from 0 to 1, where 0 indicates no functional decline and 1 indicates complete functional loss. The physician's first impression score is a clinician's subjective assessment of the patient's risk based on the initial consultation, also ranging from 0 to 1.

[0023] Laboratory biomarkers include soluble urokinase-type plasminogen activator receptor (PGAPR), mid-fraction of adrenomedullin precursor, D-dimer, C-terminal pro-vasopressin, peroxidase-4, C-reactive protein (CRP), and procalcitonin. Among these, PGAPR reflects systemic inflammatory activity, mid-fraction of adrenomedullin precursor reflects organ perfusion status, D-dimer reflects coagulation activation status, CRP reflects neuroendocrine stress response, peroxidase-4 reflects oxidative stress level, and both CRP and procalcitonin reflect the degree of inflammation. Procalcitonin is also associated with bacterial infection-related inflammation. Vital signs include heart rate, systolic blood pressure, blood oxygen saturation, respiratory rate, and body temperature.

[0024] To eliminate differences in measurement units and dimensions among various indicators, such as soluble urokinase plasminogen activator receptor measured in ng / ml and heart rate in bpm, this embodiment employs the Z-score algorithm to standardize all raw data. Specifically, standardization involves transforming the raw measurement values ​​of each indicator for the current patient based on the mean and standard deviation of each indicator in the historical training data. The historical training data must include emergency room visit data from at least 1000 elderly NSC patients, and these data must include 30-day outcome labels. This transformation ensures that all indicator data have uniform model fusion adaptability, avoiding distortion of risk contribution due to differences in dimensions.

[0025] In S2 above, this embodiment uses a preset risk mapping rule to convert each standardized indicator into a uniform-scale risk value within the range of 0 to 1, where 0 represents no risk and 1 represents extremely high risk. The risk mapping rule is designed based on the pathological significance of the indicators and the physiological characteristics of elderly patients.

[0026] In this embodiment, a specific nonlinear function form is used for the pathological mechanism of different indicators.

[0027] For the soluble urokinase-type plasminogen activator receptor, which reflects broad-spectrum inflammatory activity, a piecewise quadratic function is used:

[0028] The risk value is 0 when the index is below 4.0 ng / ml, increases twice as the value increases when it is between 4.0 and 6.5 ng / ml, and is fixed at 1.0 when it exceeds 6.5 ng / ml, thus reflecting the risk difference in different concentration ranges.

[0029] For D-dimer, which reflects the coagulation activation state, an age-dynamic correction threshold is introduced. This threshold is the patient's age divided by 100; for example, the threshold for a 75-year-old patient is 0.75 μg / ml. A logarithmic function is used for mapping.

[0030] The more the value exceeds the threshold, the more significant the increase in risk value, which is more in line with the risk characteristics of coagulation abnormalities in elderly patients.

[0031] For the mid-fraction of the adrenal medullary precursor, which reflects organ perfusion status, the Sigmoid function is used for setting:

[0032] The risk inflection point was set at 1.0 nmol / L. After exceeding this value, the risk value increased sharply, in order to reflect the strong correlation between the increase of this indicator and the risk of organ failure in clinical practice.

[0033] For ADL scores, which reflect the degree of functional decline, a power function is used to amplify the risk of patients with high functional decline by a power of 1.5:

[0034] For example, the risk value is approximately 0.465 when the ADL score is 0.6 and approximately 0.715 when the ADL score is 0.8, highlighting the impact of functional decline on risk.

[0035] For heart rate, which reflects circulatory function, a nonlinear function based on the normal range is used:

[0036] The normal range is 60 to 100 bpm, with a risk value of 0 within this range. When the risk value is below 60 bpm or above 100 bpm, the risk value increases non-linearly with the degree of deviation, which is consistent with the influence of abnormal heart rate on the circulatory stability of elderly patients.

[0037] All key parameters of the mapping functions, including inflection points, growth rates, and saturation values, are automatically optimized based on the statistical distribution of historical training data. The optimization objective is to maximize the area under the curve for single-indicator prediction of adverse outcomes over 30 days, ensuring that the parameters possess clinical rationality and data-driven scientific validity. Simultaneously, all mapping results are reduced to the 0-1 interval, achieving a unified expression of risk probability.

[0038] Taking a 75-year-old female patient as an example, the original values ​​and risk mapping results of some indicators are as follows: soluble urokinase type plasminogen activator receptor is 5.8 ng / ml, with a risk mapping value of 0.324; D-dimer is 1.5 μg / ml, with a risk mapping value of 0.084; adrenaline precursor mid-fraction is 1.2 nmol / L, with a risk mapping value of 0.645; ADL score is 0.6, with a risk mapping value of 0.465; heart rate is 110 bpm, with a risk mapping value of 0.477. These values ​​provide a clear understanding of the risk of individual indicators.

[0039] The risk values ​​obtained in S2 will be used as inputs in the next step of risk sub-model construction for multi-dimensional disease mechanism aggregation analysis.

[0040] For S3, this step divides the individual risk values ​​obtained in step S2 into five risk sub-models according to the pathological mechanism dimension. The dimensional risk score of each sub-model is calculated by weighted linear combination. The dimensional risk score ranges from 0 to 1, realizing the structured attribution of risk sources and improving the interpretability of risk assessment.

[0041] The five risk sub-models are categorized based on the pathological mechanisms of biomarkers. The inflammation-immunity sub-model integrates individual risk values ​​related to inflammation and immunity, with core indicators including soluble urokinase-type plasminogen activator receptor, C-reactive protein, procalcitonin, and peroxidase-4, primarily reflecting the patient's systemic inflammation and immune activation status. The coagulation-circulation sub-model integrates individual risk values ​​related to coagulation and cardiovascular function, with core indicators including D-dimer, systolic blood pressure, and heart rate, reflecting the patient's coagulation abnormalities and circulatory stability. The organ function sub-model integrates individual risk values ​​related to organ perfusion and respiratory function, with core indicators including adrenaline precursor mid-fraction, blood oxygen saturation, and respiratory rate, reflecting the patient's risk of organ damage and functional failure. The clinical perception sub-model integrates individual risk values ​​related to clinical subjective assessment and functional status, with core indicators including physician first impression score and ADL score, reflecting clinically apparent risk and functional reserve. The basic risk moderating sub-model uses age as the core indicator to reflect the basic risk of elderly patients due to physiological degeneration. The risk increases quadratically after age 65, and the risk value is 0 when the age is 65 or below.

[0042] The weights of the indicators within each sub-model are calculated based on historical training data. The calculation criteria include the area under the curve of a single indicator, the logistic regression coefficient, or the importance of gradient boosting tree features. Furthermore, the sum of the weights of all indicators within each sub-model is 1, ensuring the rationality and comparability of the weights.

[0043] In a feasible scenario, in the inflammation-immune sub-model, the weights of soluble urokinase plasminogen activator receptor are 0.35, C-reactive protein is 0.30, procalcitonin is 0.20, and peroxidase-4 is 0.15. In the coagulation-circulation sub-model, the weight of D-dimer was 0.50, systolic blood pressure was 0.20, and heart rate was 0.30; in the organ function sub-model, the weight of the mid-fraction of adrenal medullary precursor was 0.55, blood oxygen saturation was 0.25, and respiratory rate was 0.20. In the clinical perception sub-model, the weight of the doctor's first impression score is 0.55, and the weight of the ADL score is 0.45. In the basic risk adjustment sub-model, age has a weight of 1.00, and its dimension score is calculated separately.

[0044] The dimensional risk score for each sub-model is the weighted sum of all individual risk values ​​and their corresponding weights within that sub-model. Taking the inflammation-immunity sub-model as an example, the dimensional score is...

[0045] The coagulation-circulation sub-model dimension score is:

[0046] The organ function sub-model dimension score is:

[0047] The scores for the clinical perception sub-model dimensions were obtained by aggregating the following indicators: the degree of decline in ADL (Awake Daily Living) scores with the Doctor's Initial Impression score DOI:

[0048] The basic risk adjustment sub-model is set as follows:

[0049] In the above model, all Risk mapping results from S2; all Weights of internal metrics obtained during training of the sub-model; Sub-model output value This indicates the relative risk contribution under this pathological mechanism dimension.

[0050] This step still uses a 75-year-old female patient as an example, assuming her C-reactive protein risk value is 0.6, procalcitonin risk value is 0.5, peroxidase-4 risk value is 0.4, systolic blood pressure risk value is 0.3, blood oxygen saturation risk value is 0.2, respiratory rate risk value is 0.3, and doctor's first impression score risk value is 0.7. Combining these with the previous individual risk values, the scores for each sub-model dimension can be calculated: the inflammation-immunity sub-model dimension score is 0.35×0.324+0.30×0.6+0.20×0.5+0. 15×0.4=0.6275; Coagulation-circulation sub-model dimension score is 0.50×0.084+0.20×0.3+0.30×0.477=0.69; Organ function sub-model dimension score is 0.55×0.645+0.25×0.2+0.20×0.3=0.64; Clinical perception sub-model dimension score is 0.55×0.7+0.45×0.465=0.655; Basic risk regulation sub-model dimension score is 0.02×(75-65)²=3.38.

[0051] In S4, this embodiment sets up a two-layer weighted fusion mechanism to integrate the interaction effect of the dimensional scores of the five sub-models with the individual risk values ​​to obtain the final total risk score.

[0052] Specifically, first, set up the overall scoring function:

[0053] in The scoring results for the five sub-models; The global weights of the kkk-th sub-model (derived from training optimization); Standardized individual risk indicators used in the scoring; The penalty weights between index pairs known to have a reinforcing effect; The intensity of interaction factor adjustment (hyperparameter, which adjusts the degree of influence of the interaction term weights on F(X)).

[0054] Sigmoid risk score function:

[0055] The RiskScore output range of this function is (0,1), which represents the overall risk probability and can be used for three-stage classification (low / medium / high risk).

[0056] Based on the aforementioned prerequisites, the first layer of the fusion strategy in this embodiment is a global weighting of sub-models. Each sub-model is assigned a global weight based on its contribution to the prediction of adverse outcomes over 30 days. These global weights are obtained by training the complete model using logistic regression or gradient boosting tree algorithms, and are normalized based on the predictive importance of the sub-models. The sum of the global weights of all sub-models is 1. For example, the global weight of the inflammation-immune sub-model is 0.25, the coagulation-circulation sub-model is 0.20, the organ function sub-model is 0.20, the clinical perception sub-model is 0.15, and the basic risk regulation sub-model is 0.20. These weights reflect the differences in the contribution of different sub-models to the overall risk.

[0057] The second layer of the fusion strategy is interaction term penalty, which corrects for individual risk pairs exhibiting a synergistic worsening effect to avoid underestimating risk. A synergistic worsening effect refers to a situation where the risk level is significantly higher when some indicators simultaneously increase than when they increase individually, such as when soluble urokinase-type plasminogen activator receptor (USPR) and C-reactive protein (CRP) simultaneously increase, the inflammatory risk is significantly higher than when either increases individually. Clinically proven synergistic effect indicator pairs are selected, including those mentioned above, such as the mid-fraction of adrenaline precursor and respiratory rate, and blood oxygen saturation and systolic blood pressure. Interaction weights are determined through interaction variable regression or SHAP interaction analysis, ranging from -1 to 1. Positive values ​​indicate increased synergistic risk, while negative values ​​indicate decreased synergistic risk. The default weight for high-risk interaction pairs, such as USPR and CRP, is 0.25, and for weakly correlated interaction pairs, such as blood oxygen saturation and systolic blood pressure, it is 0.10. Simultaneously, the interaction factor regulation strength is set; this is a hyperparameter with a default value of 1.0, used to control the influence of interaction terms on the total score.

[0058] In this embodiment, the calculation of the total risk score is divided into three steps. The first step is to calculate the linear weighted sum of the scores for the five sub-model dimensions, i.e., the sum of the scores for each sub-model dimension multiplied by its corresponding global weight. The second step is to calculate the interaction term correction value, i.e., the sum of the interaction factor adjustment strength multiplied by the interaction weight of each interaction pair and the product of the two individual risk values ​​in that interaction pair, retaining only the top few interaction pairs that significantly contribute to risk prediction. The third step is to normalize the sum of the linear weighted sum and the interaction term correction value using the Sigmoid function to obtain the total risk score. This function maps the value to between 0 and 1, eliminating numerical biases in different calculation stages and ensuring the stability and comparability of the total risk score.

[0059] This embodiment continues to use a 75-year-old female patient as an example. The first step is to calculate the linear weighted sum: 0.25×0.6275+0.20×0.69+0.20×0.64+0.15×0.655+0.20×3.38=1.19715. The second step is to calculate the interaction term correction value. Assuming the interaction factor regulation strength is 1.0, three interaction pairs are selected: soluble urokinase-type plasminogen activator receptor with C-reactive protein, adrenal medullary precursor mid-fraction with respiratory rate, and blood oxygen saturation with systolic blood pressure. Their interaction weights are 0.25, 0.15, and 0.10, respectively. The corresponding individual risk value products are 0.324×0.6, 0.645×0.3, and 0.2×0.3, respectively. The interaction term correction value is 1.0×(0.25×0.324×0.6+0.15×0.645×0.3+0.10×0.2×0.3)=0.1875. The third step is to calculate the total risk score: 1÷(1+e^(-(1.19715+0.1875)))≈0.80. The patient's total risk score is 0.80, which falls into the high-risk category.

[0060] Through the above S1-S4, this embodiment obtains the patient's total risk score. Therefore, in S5, this step converts the total risk score into an operable clinical risk level and links with the emergency information system to push treatment pathway suggestions, realizing closed-loop management from risk assessment to clinical intervention, and improving the efficiency of emergency treatment and patient prognosis.

[0061] As an optional approach, risk grading employs a dual-threshold, three-tiered classification method, with baseline thresholds determined based on ROC curve analysis of historical data. Low risk corresponds to a total risk score less than 0.35, with a typical probability of adverse outcome within 30 days of less than 5%; medium risk corresponds to a total risk score between 0.35 and 0.65, with a typical probability of adverse outcome within 30 days of 5% to 20%; and high risk corresponds to a total risk score greater than or equal to 0.65, with a typical probability of adverse outcome within 30 days of more than 20%.

[0062] To adapt to the characteristics of patient populations in different hospitals, such as the differences in risk distribution between community hospitals and tertiary hospitals, the model needs to be retrained and the grading thresholds updated quarterly based on real data and 30-day outcomes of elderly NSC patients in our hospital. The update process is as follows: First, assessment data and 30-day outcomes (including death, hospitalization, and ICU admission) of at least 200 elderly NSC patients in our hospital are collected quarterly. Then, the existing model structure is used to refit the new data, and the ROC curve is calculated. Next, the score at which the Youden index is maximized is selected as the new threshold. The Youden index is calculated by adding sensitivity and specificity minus 1, which can balance sensitivity and specificity. Finally, the new threshold is hot-loaded into the system through the model configuration file, and it takes effect without restarting the system, ensuring the timeliness and adaptability of the threshold.

[0063] Output corresponding clinical diagnosis and treatment path suggestions based on the risk level, with different emphases for suggestions of different risk levels. For low-risk patients, outpatient assessment or home follow-up suggestions are pushed. The system automatically generates a follow-up template, which includes a 72-hour phone follow-up reminder and medical guidance when symptoms worsen. Such patients do not need to be hospitalized or observed, reducing unnecessary occupation of medical resources. For medium-risk patients, suggestions for observation and reexamination are pushed, indicating to recheck core biomarkers such as soluble urokinase-type plasminogen activator receptor and mid-regional proadrenomedullin fragment after 6 to 12 hours. At the same time, primary diagnosis is linked with secondary departments such as geriatrics to closely monitor the patient's condition changes. For high-risk patients, a red alert is pushed, and the system automatically triggers multiple tasks:置顶显示患者信息 to the duty doctor's terminal to ensure that doctors pay priority attention; link with the ICU consultation system to queue the patient into the priority consultation queue to shorten the consultation waiting time; dock with the bed management system to pre-allocate an ICU or emergency observation bed to avoid delays caused by bed shortages; generate an emergency admission process sheet, which includes examination items that need to be completed first, such as blood gas analysis and chest CT, to speed up the admission disposal speed.

[0064] Taking a 75-year-old female patient as an example, her total risk score is 0.80, belonging to high risk. The linkage operations triggered by the system include: a pop-up window of a high-risk patient reminder is displayed on the emergency information system duty main console, suggesting an immediate ICU consultation; an SMS notification is automatically sent to the ICU duty doctor, and the core index information of the patient is attached to the SMS, such as the soluble urokinase-type plasminogen activator receptor is 5.8 ng / ml and the mid-regional proadrenomedullin fragment is 1.2 nmol / L; the bed system shows that there is 1 available ICU bed and locks the bed for 30 minutes to reserve time for the patient; after the doctor confirms, the system automatically generates an ICU admission assessment form, reducing the manual document processing time and improving the overall disposal efficiency.

[0065] In addition, to ensure the generalization ability and clinical adaptability of the model, this embodiment also systematically trains and optimizes all parameters of the model through historical data. The parameters include the index weights within the sub-model, the global weights of the sub-model, and the interaction term weights. The training process focuses on the combination of data-driven and clinical rationality.

[0066] It should be noted that there is an unclear expression "置顶显示患者信息" in the translation of item , which needs to be further clarified according to the specific meaning to ensure the accuracy of the translation.In one optional scenario, the training data must meet certain requirements to ensure training effectiveness. Regarding sample size, it must include at least 1000 NSC emergency room patients aged 65 and above, and the sample must cover patients of different genders and with different underlying diseases such as hypertension and diabetes to ensure representativeness. Regarding data integrity, each sample must include all core indicators from step S1, including raw and standardized values, to avoid affecting model training due to missing indicators. Regarding outcome labels, each sample must have a clear 30-day outcome label, including at least one of the following: death, hospitalization, or ICU admission, providing a clear training objective for the model. Regarding data sources, historical case data should be extracted from hospital information systems, laboratory information systems, and electronic medical records, excluding samples with a missing indicator rate exceeding 20% ​​to ensure data quality.

[0067] Parameter training employs a targeted strategy. For training weights within sub-models, each sub-model is modeled independently, using either logistic regression or gradient boosting tree algorithms. Logistic regression uses the regression coefficient as the weight, with larger coefficients indicating a greater contribution to risk prediction. Gradient boosting trees use feature importance, with higher feature importance indicating a more critical role for the indicator in the model. The original weights are then normalized to ensure the sum of all indicator weights within each sub-model is 1. For training global weights of sub-models, the dimensional scores of the five sub-models are used as input variables to construct a logistic regression model predicting adverse outcomes over 30 days. The global weights are obtained by normalizing the regression coefficients of each sub-model's dimensional scores, reflecting the overall contribution of each sub-model to overall risk prediction. For training interaction term weights, indicator pairs are added as new variables to the model through interaction variable regression. The regression coefficients of these new variables are used as interaction term weights, with weight values ​​constrained between -1 and 1. Only interaction pairs with a p-value less than 0.05 in the statistical test are retained to ensure the statistical significance of the interaction terms' contribution to risk prediction.

[0068] In the model evaluation and optimization phase, this embodiment employs multiple metrics to comprehensively assess model performance, avoiding the limitations of a single metric. The area under the curve (AUC) is used to evaluate the model's discriminative ability; an AUC exceeding 0.85 indicates excellent discriminative ability. The F1 score balances precision and recall; an F1 score exceeding 0.8 indicates excellent performance in both precision and recall. The Brier score assesses the calibration of risk probabilities; a Brier score below 0.1 indicates excellent calibration. To avoid overfitting, a 5-fold cross-validation method is used, dividing the training data into five parts. Four parts are used as the training set and one part as the validation set in each iteration, repeated five times, and the average performance metric is calculated. In terms of algorithm optimization, grid search and early stopping strategies are used to optimize hyperparameters such as the interaction factor adjustment strength. Grid search finds the optimal value by traversing the preset parameter combinations, and early stopping stops training when the model performance no longer improves to prevent overfitting. At the same time, L1 regularization, i.e., LASSO, is used to screen redundant indicators. For example, peroxide reductase-4 is removed if the area under the curve of a single indicator is less than 0.65, which simplifies the model structure while ensuring prediction performance.

[0069] Through the above approach, this embodiment achieves precise risk stratification for elderly NSC emergency patients. Compared to existing technologies, this embodiment covers multiple dimensions such as inflammation, coagulation, and organ function through multi-marker fusion, integrating novel markers such as soluble urokinase-type plasminogen activator receptor and adrenal medullary precursor mid-fragment to improve prediction accuracy. It fully adapts to the physiological characteristics of elderly patients through age-corrected thresholds and age-based risk sub-models. It boasts strong clinical operability, with the system supporting automatic data access and path pushing, reducing manual intervention and improving emergency treatment efficiency. It also possesses dynamic adaptability, allowing hospitals to update thresholds and weights based on local data, ensuring the model's applicability in different clinical scenarios. It can be directly applied to emergency geriatric medicine departments in hospitals at all levels, helping doctors identify high-risk patients early, optimize medical resource allocation, and improve patient prognosis.

[0070] This embodiment also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of an emergency risk stratification method for elderly NSC patients.

[0071] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of a method for risk stratification in emergency care for elderly patients with NSC.

[0072] The embodiments of the present invention have been described in detail above. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A risk stratification method for elderly patients with NSC emergency care, characterized in that, include: S1: Obtain multidimensional assessment data of elderly NSC patients and standardize the multidimensional assessment data; S2: For each standardized assessment data point, a preset risk mapping rule is used to convert it into a single risk value with a uniform scale; the risk mapping rule is set according to the pathological significance of each assessment data point and the physiological characteristics of elderly NSC patients; S3: Divide the single risk value into multiple risk sub-models, and calculate the dimensional risk score for each risk sub-model; S4: Based on a preset fusion strategy, the dimensional risk scores of multiple risk sub-models and the interaction effects between the individual risk values ​​are integrated to calculate the patient's total risk score. S5: Determine the patient's risk level based on the total risk score, and output corresponding clinical treatment pathway recommendations based on the risk level.

2. The method for risk stratification in emergency care for elderly patients with NSC according to claim 1, characterized in that, The multidimensional assessment data includes at least basic clinical information, laboratory biomarker indicators, and vital sign indicators. The laboratory biomarker indicators include at least one of the following: soluble urokinase type plasminogen activator receptor, adrenal medullaris precursor mid-terminus, D-dimer, angiotensin C-terminal precursor, and peroxidase-4.

3. The method for risk stratification in emergency care for elderly patients with NSC according to claim 1, characterized in that, The standardization process employs the Z-score algorithm, which achieves standardization by eliminating differences in measurement units and dimensions among different evaluation data. This includes: Based on the mean and standard deviation of each assessment data in the historical training data, the original values ​​of the current patient's assessment data are transformed and standardized.

4. The method for risk stratification in emergency care for elderly patients with NSC according to claim 1, characterized in that, The preset risk mapping rules include nonlinear mapping functions, and adopt specific function forms for the pathological mechanisms of different assessment data; Among them, a piecewise quadratic function is used for indicators reflecting broad-spectrum inflammatory activity, a sigmoid function is used for indicators reflecting organ perfusion status, a power function is used for indicators reflecting the degree of functional decline, and a nonlinear function based on the normal range is used for indicators reflecting heart rate.

5. The method for risk stratification in emergency care for elderly patients with NSC according to claim 4, characterized in that, The risk mapping rule also includes an age-based dynamic correction threshold, which is used to distinguish between normal and abnormal risk ranges for D-dimer.

6. The method for risk stratification in emergency care for elderly patients with NSC according to claim 1, characterized in that, The output range of the single risk value of the unified scale is set with a preset interval. The values ​​within the preset interval are used to realize the probabilistic expression of the risk of the single assessment data, and the key parameters of each risk mapping rule are automatically optimized based on the statistical distribution of historical training data.

7. The method for risk stratification in elderly patients with NSC emergency care according to claim 1, characterized in that, The multiple risk sub-models include five categories of sub-models based on pathological mechanisms: inflammation-immunity sub-model, coagulation-circulation sub-model, organ function sub-model, clinical perception sub-model, and basic risk regulation sub-model. Among them, the inflammation-immunity sub-model is used to integrate individual risk values ​​related to inflammation and immunity; the coagulation-circulation sub-model is used to integrate individual risk values ​​related to coagulation and cardiovascular function; the organ function sub-model is used to integrate individual risk values ​​related to organ perfusion and respiratory function; the clinical perception sub-model is used to integrate individual risk values ​​related to the degree of functional decline and the doctor's initial impression; and the basic risk regulation sub-model is constructed based on the patient's age.

8. The method for risk stratification in elderly patients with NSC emergency care according to claim 7, characterized in that, The dimensional risk score of each risk sub-model is a weighted linear combination of all individual risk values ​​under that sub-model; The weight of each individual risk value is calculated based on historical training data, and the sum of the weights of all individual risk values ​​within each sub-model is 1.

9. The method for risk stratification in emergency care for elderly patients with NSC according to claim 1, characterized in that, The preset fusion strategy includes: The first fusion layer is set as a global weighted sub-model, and the dimensional risk scores of the five risk sub-models are linearly aggregated through global weights; The second fusion layer is set as an interaction item penalty, and the individual risk value pairs with synergistic deterioration effect are corrected by interaction weight. The individual risk value pairs with synergistic deterioration effect include one or more of the following: suPAR and CRP, MR-proADM and respiratory rate, blood oxygen saturation and systolic blood pressure.

10. The method for risk stratification in emergency care for elderly patients with NSC according to claim 1, characterized in that, The risk levels include three levels: low risk, medium risk, and high risk. Each level has a different total risk score range. The threshold of the score range is dynamically updated based on the hospital's local historical data. The update method is to determine the optimal threshold by calculating the Youden index through ROC curve analysis every quarter.

11. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of a method for risk stratification in emergency care for elderly patients with NSC as described in any one of claims 1-7 when executing the computer program.

12. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for risk stratification in emergency care for elderly patients with NSC as described in any one of claims 1-7.

Images