Postoperative sub-delirium syndrome risk prediction system for cardiac surgery patient

Abstract

The invention discloses a postoperative sub-delirium syndrome risk prediction system for a cardiac surgery patient, and relates to the technical field of medical treatment, and the system comprises a clinical data collection module which is used for collecting multi-dimensional clinical data of the cardiac surgery patient in a perioperative period under extracorporeal circulation, the data comprises patient baseline information, preoperative neurocognitive function evaluation indexes, intraoperative physiological parameters and postoperative early-stage vital signs; and the risk identification module is used for utilizing a single-factor statistical analysis mechanism. According to the postoperative sub-delirium syndrome risk prediction system for the cardiac surgery patient, accurate prediction of the sub-delirium syndrome risk is achieved by collecting multi-dimensional clinical data of the patient in the perioperative period, such as baseline information, preoperative neurocognitive function, intraoperative physiological parameters and postoperative early-stage vital signs; the independent risk factors are scientifically screened and quantified by using single factor analysis and a multivariate Logistic regression model, and the problem of incomplete data dimensions in traditional evaluation is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of medical technology, specifically to a risk prediction system for postoperative delirium syndrome in cardiac surgery patients. Background Technology

[0002] Hypnotic delirium syndrome, a common postoperative neurocognitive disorder in patients undergoing cardiopulmonary bypass surgery, is characterized by decreased clarity of consciousness, poor concentration, and fluctuating cognitive function. Although less severe than delirium, it is still closely associated with delayed postoperative recovery, prolonged hospital stay, and an increased risk of long-term cognitive impairment. Clinical practice shows that the incidence of hypodelirium in this population can reach 20%-40%, and its occurrence is influenced by multiple factors, including preoperative baseline condition, intraoperative physiological disturbances, and postoperative inflammatory response. Therefore, accurate identification of high-risk patients and early intervention are of significant clinical value in improving postoperative prognosis.

[0003] Currently, risk assessment for postoperative delirium in cardiac surgery patients largely relies on single indicators or empirical judgment. While some studies have attempted to incorporate limited clinical variables, they suffer from incomplete data coverage and insufficient consideration of multifactorial interactions. Furthermore, existing predictive models often lack systematic statistical validation and dynamic updating mechanisms, making it difficult to adapt to the characteristics and differences among patient groups in different medical centers. Consequently, their accuracy and applicability in clinical application require further improvement. To address these issues, we propose a risk prediction system for postoperative delirium syndrome in cardiac surgery patients. Summary of the Invention

[0004] To address the aforementioned technical issues, a risk prediction system for postoperative delirium syndrome in cardiac surgery patients is provided. This technical solution resolves the problems mentioned above regarding risk assessment of postoperative delirium in cardiac surgery patients, which often relies on single indicators or empirical judgments, has incomplete data coverage, and does not adequately consider the interaction of multiple factors. Furthermore, the prediction model lacks systematic statistical validation and dynamic update mechanisms, making it difficult to adapt to differences among different patient groups, and its accuracy and applicability need to be improved.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A risk prediction system for postoperative delirium syndrome in cardiac surgery patients includes: The clinical data acquisition module is used to collect multidimensional clinical data of patients undergoing cardiac surgery under cardiopulmonary bypass during the perioperative period. The data includes patient baseline information, preoperative neurocognitive function assessment indicators, intraoperative physiological parameters, and early postoperative vital signs. The risk identification module is used to screen candidate risk factors that are significantly associated with delirium syndrome from the collected multidimensional clinical data using a univariate statistical analysis mechanism. The risk assessment quantification module is used to determine independent risk factors and their degree of influence by inputting candidate risk factors into a multivariate logistic regression model and based on the backward stepwise regression method. The risk score generation module is used to calculate the patient's individualized total risk score based on the independent risk factors identified in the assessment and their degree of influence. The risk probability calculation module is used to convert the total risk score into the incidence probability value of delirium syndrome.

[0006] Preferably, the clinical data acquisition module acquires multidimensional clinical data specifically as follows: The patient's baseline information includes, but is not limited to: age, gender, years of education, history of preoperative cognitive impairment, history of use of psychotropic drugs, and history of alcoholism; The preoperative neurocognitive function assessment indicators include, but are not limited to: Montreal Cognitive Assessment Scale score, Hamilton Anxiety Rating Scale score, and EEG alpha wave power percentage. The intraoperative physiological parameters include, but are not limited to: duration of cardiopulmonary bypass, maximum decrease in cerebral oxygen saturation from baseline, mean arterial pressure fluctuation coefficient, and cumulative remifentanil dosage. The early postoperative vital signs include, but are not limited to: Richmond agitation-sedation score, interleukin-6 serum concentration, and number of sleep-wake cycle changes within 24 hours.

[0007] Preferably, the risk identification module uses the following method to screen candidate risk factors: We collected perioperative multidimensional clinical data from all patients undergoing cardiac surgery under cardiopulmonary bypass, preprocessed the multidimensional clinical data, and used the occurrence of delirium syndrome as a binary outcome variable. Perform independent statistical analysis for each clinical variable, and select the statistical test method according to the variable type: For continuous variables that conform to a normal distribution, the independent samples t-test is used; for continuous variables that do not conform to a normal distribution, the Mann-Whitney U test is used; the continuous variables include, but are not limited to: age, Montreal Cognitive Assessment Scale score, and the maximum decrease in brain oxygen saturation from baseline; Chi-square test was used for categorical variables; the categorical variables include, but are not limited to: gender, history of preoperative cognitive impairment, and history of alcoholism. Set a significance level threshold, calculate the p-value for each variable test. If the p-value is less than the significance level threshold, it indicates that the variable is statistically significantly associated with delirium syndrome and is included in the candidate risk factor set. All variables with significant p-values ​​were summarized as candidate risk factors and used as input for the subsequent multivariate logistic regression model.

[0008] Preferably, the risk assessment quantification module determines independent risk factors and their degree of influence using the following method: All candidate risk factors identified through univariate analysis were incorporated into the initial multivariate logistic regression model, and the iterative calculation process was initiated: In each step, a likelihood ratio test is performed on all variables in the current model, and the variable with the largest p-value that is higher than the preset removal threshold is removed, with only one variable removed at a time. Repeat this elimination process until the p-values ​​of all remaining variables are less than the significance level threshold; After each removal, the model is refitted and the model fit is verified. The variables that are finally retained are the independent risk factors. For each identified independent risk factor, its adjusted odds ratio (OR) and confidence interval were calculated. The OR value represents the degree of influence of the factor on the risk of developing delirium syndrome, and the regression coefficient β value was recorded for subsequent scoring system construction.

[0009] Preferably, the method for determining independent risk factors and their degree of influence requires collinearity diagnosis throughout the entire process, specifically as follows: After obtaining the final set of independent risk factors through backward stepwise regression, a linear regression model containing all identifiable factors is constructed. For each independent variable Xᵢ in the model, all other independent variables are used as predictors and Xᵢ is used as the dependent variable to establish an auxiliary linear regression model, and the determination coefficient R²ᵢ of the auxiliary model is calculated. Based on the calculated coefficient of determination, calculate the variance inflation factor of the independent variable; After completing the above calculations for all independent variables, check the VIF value of each variable: if all VIF values ​​are less than the preset value, it is determined that there is no significant multicollinearity; if there is a variable with VIF ≥ the preset value, identify the variable and assess its clinical significance, and delete the variable. After processing, rebuild the regression model and repeat the above VIF calculation process until the VIF values ​​of all variables in the final model are stable below the preset value; The final output consists of independent risk factors that pass the collinearity test and their regression coefficients.

[0010] Preferably, the risk score generation module calculates the patient's individualized total risk score using the following method: Find the minimum value among all independent risk factor regression coefficients, calculate the weight value of each factor, which is equal to the regression coefficient of the factor divided by the minimum regression coefficient, and round the result to the nearest integer. The total risk score is obtained by summing the products of the weights of all independent risk factors and their indications of presence or absence, where the indications are determined according to the risk factor determination criteria preset by the system.

[0011] Preferably, the method for converting the total risk score into the incidence probability value of delirium syndrome is as follows: Receive the patient's individualized total risk score from the risk score generation module, and substitute the total risk score into the risk prediction function for calculation; The risk prediction function is an S-shaped curve function, and its output value represents the probability of disease occurrence; Key parameters in the function include: the slope parameter that controls the steepness of the curve and the cutoff threshold that determines the risk threshold. The values ​​of the key parameters are determined by applying the maximum likelihood estimation method to the training set data for optimization and fitting.

[0012] Preferably, the slope parameter and the cutoff threshold are determined by plotting the subject operating characteristic curve on the validation set and selecting the parameter combination that maximizes the area under the curve and has the highest Youden index.

[0013] Preferably, the training set data specifically includes: Complete perioperative data of patients undergoing cardiopulmonary bypass surgery in the past were collected during the system development phase. The data included all dimensions, including: patient baseline information, preoperative neurocognitive assessment, intraoperative physiological parameters, and early postoperative vital signs. The key outcome label included whether the patient developed delirium syndrome postoperatively.

[0014] Preferably, it also includes a dynamic update module, which connects the risk grading and intervention module, specifically: Real-time reception of consciousness status assessment results based on the CAM-ICU scale within 72 hours postoperatively; When the cumulative number of newly added delirium events is ≥5, the model is retrained, the newly added data is added to the training set, and the single factor analysis of the risk identification module and the multiple regression of the risk assessment quantification module are re-executed. Update the weight values ​​of the risk scoring generation module and the slope parameters and cutoff threshold of the risk probability calculation module.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The proposed system for predicting the risk of postoperative delirium syndrome in cardiac surgery patients utilizes multidimensional clinical data collected during the perioperative period, such as baseline information, preoperative neurocognitive function, intraoperative physiological parameters, and early postoperative vital signs. This enables accurate prediction of the risk of delirium syndrome. By employing univariate analysis and multivariate logistic regression models, it scientifically screens and quantifies independent risk factors, effectively addressing the problems of incomplete data dimensions and insufficient consideration of multifactor interactions in traditional assessments. The system also features dynamic updating capabilities, automatically adjusting model parameters based on new data to ensure the accuracy and timeliness of predictions. It converts the total risk score into a probability value, providing clinicians with intuitive and easy-to-use decision support. This helps in the early identification and intervention of high-risk patients, thereby improving postoperative prognosis and reducing the medical burden. Attached Figure Description

[0016] Figure 1 This is a system module diagram of the present invention; Figure 2 This is a system workflow diagram of the present invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 and 2As shown, the risk prediction system for postoperative delirium syndrome in cardiac surgery patients undergoing cardiopulmonary bypass is designed to transform complex perioperative clinical information into quantifiable individual risk prediction values ​​through a systematic, multi-stage analysis process. The system begins with a clinical data acquisition module, responsible for comprehensively collecting multidimensional perioperative data from patients. In the patient baseline information section, age is considered an important factor as it is associated with decreased brain reserve function; gender differences may affect neuroendocrine responses; years of education serve as a proxy indicator of cognitive reserve; a history of preoperative cognitive impairment directly suggests neurological vulnerability; a history of psychotropic drug use may interfere with neurotransmitter balance; and a history of alcohol abuse is associated with postoperative withdrawal risk. For preoperative neurocognitive function assessment, the Montreal Cognitive Assessment Scale provides a standardized overall cognitive function screening; the Hamilton Anxiety Rating Scale quantifies anxiety state, as anxiety is known to be a triggering factor for delirium; and the proportion of alpha wave power on electroencephalogram (EEG) serves as an objective indicator of brain function activity, reflecting the balance between cortical excitation and inhibition. Intraoperative physiological parameter acquisition focuses on key aspects: the duration of cardiopulmonary bypass is directly related to the duration of cerebral hypoperfusion exposure; the maximum decrease in cerebral oxygen saturation from baseline is monitored in real time to detect the imbalance between cerebral oxygen supply and demand, and this indicator is continuously acquired using near-infrared spectroscopy; the mean arterial pressure fluctuation coefficient is calculated by dividing the standard deviation by the mean to quantify hemodynamic instability; remifentanil, as a highly selective μ-receptor agonist, requires precise recording of its cumulative dosage due to the definite neurotoxicity of opioids. Early postoperative monitoring includes the Richmond agitation-sedation score, used to objectively assess consciousness and sedation depth; interleukin-6, as a representative of pro-inflammatory cytokines, shows elevated serum concentrations postoperatively, indicating the intensity of systemic inflammatory response; the number of sleep-wake cycle transitions within 24 hours is obtained through motion recording or electroencephalography (EEG) monitoring, reflecting the degree of circadian rhythm disruption. The selection of these indicators is not arbitrary; each has a solid pathophysiological basis, corresponding to the core mechanisms of delirium: neuroinflammation, neurotransmitter imbalance, cerebral perfusion injury, and circadian rhythm dysregulation.

[0019] The core task of the risk identification module is to initially screen candidate factors significantly associated with hypodelusion from a large number of clinical variables. During the data preprocessing stage, rigorous missing value handling, outlier correction, and variable distribution testing are required. After defining the occurrence of hypodelusion syndrome as a binary outcome variable, the module initiates a univariate statistical analysis process. The selection of analytical methods strictly follows variable characteristics and statistical principles: for continuous variables conforming to a normal distribution, such as age, Montreal Cognitive Assessment Scale score, and maximum decrease in cerebral oxygen saturation, the independent samples t-test (IPS) is used to compare the mean differences between the delirium group and the non-delirium group; for continuous variables that do not conform to a normal distribution, the nonparametric Mann-Whitney U test is used for intergroup comparisons; for categorical variables such as gender, preoperative cognitive impairment history, and alcohol abuse history, the chi-square test or Fisher's exact test (when the expected frequency is too low) is used to analyze their distribution differences. After setting a pre-determined significance level threshold, the p-value for the association test is calculated independently for each variable. If the p-value of a variable is strictly less than this threshold, it indicates that the variable has a statistically significant correlation with the delirium outcome, and it is included in the set of candidate risk factors. This step is essentially a preliminary screening, designed to efficiently eliminate a large number of irrelevant variables, focusing on a set of factors with potential predictive value, providing reasonable input for subsequent more refined multifactor modeling, and avoiding overcomplication of the model.

[0020] Candidate risk factors that pass the initial screening are then analyzed in depth within the risk assessment quantification module. The core of this module is constructing a multivariate logistic regression model and using a backward stepwise regression method for variable selection. The initial model includes all candidate variables with significant p-values ​​in the univariate analysis. The iterative calculation process is then rigorously implemented: in each iteration, a comprehensive likelihood ratio test is performed on all variables included in the current model, identifying and removing the variable with the weakest statistical significance (i.e., the largest p-value) that is above a preset removal threshold. Only one variable is removed in each iteration to ensure process stability and controllability. The model is immediately refitted after variable removal, and the Hosmer-Lemeshow test or information criterion is used to assess the model's goodness of fit. This removal-refitting cycle continues until all remaining variables in the model have p-values ​​less than a preset significance level threshold, indicating that their contribution to the outcome is statistically independent. The final set of retained variables constitutes the independent risk factors, and their impact on the risk of delirium is the net effect after excluding confounding factors. For each identified independent risk factor, its adjusted odds ratio and 95% confidence interval are precisely calculated. The odds ratio directly quantifies the increased risk of developing delirium when the factor is present. Simultaneously, the regression coefficient β value for each factor is accurately recorded; this coefficient directly determines the relative weight of the factor in the subsequent risk scoring system. To ensure the statistical reliability and interpretability of the final model, strict multicollinearity diagnosis must be performed after obtaining the set of independent risk factors. The specific implementation process is as follows: A linear regression model containing all ultimately identified factors is constructed. For each independent variable Xᵢ in this model, all other independent variables are used as predictors, and Xᵢ itself is used as the dependent variable to establish an auxiliary linear regression model, and the determination coefficient R²ᵢ of this auxiliary model is calculated. Based on the calculated R²ᵢ, the variance inflation factor (VIF) of the independent variable Xᵢ is calculated. After completing the above calculations for all independent variables, the VIF value of each variable is checked one by one. If all VIF values ​​are below the preset critical value, the model is considered to have no significant multicollinearity problem. If a variable is found to have a VIF value exceeding a preset threshold, a comprehensive evaluation considering its clinical importance and theoretical significance is necessary. Variables with relatively minor clinical significance or weak theoretical basis are typically removed. After removal, the regression model must be reconstructed, and the VIF calculation process described above repeated until the VIF values ​​of all retained variables in the final model are stably controlled below the preset threshold. Only independent risk factors that pass this rigorous collinearity test and their corresponding regression coefficients are ultimately confirmed as valid and output to the next module.

[0021] The core function of the risk score generation module is to transform abstract statistical regression coefficients into an integer risk score system that is easy to understand and apply in clinical practice. The transformation process follows standardized steps: First, the minimum value among all identified independent risk factor regression coefficients is identified. Then, a weight value for each factor is calculated, which is equal to the factor's own regression coefficient divided by the previously found minimum regression coefficient. The result is typically rounded to the nearest integer to obtain a clinically usable integer weight. The patient's individualized total risk score is obtained by summing the products of the weight values ​​of all independent risk factors and their indications of presence or absence. The determination of the indication value (usually 0 or 1) strictly depends on pre-defined risk factor judgment criteria based on clinical expertise and data distribution characteristics. For example, age may have a threshold value, with an indication value of 1 representing age ≥70 years; the maximum decrease in cerebral oxygen saturation may have a percentage threshold, with an indication value of 1 representing a decrease exceeding 30% of the baseline value. This total risk score intuitively quantifies the cumulative burden of multiple risk factors the patient is exposed to.

[0022] The risk probability calculation module is responsible for converting the total risk score into a more clinically understandable probability value for disease incidence. This module receives the individualized total risk score from the previous module and substitutes it into a pre-defined risk prediction function for calculation. This function typically uses an S-curve, with its output value directly mapped to an estimated probability of disease incidence between 0 and 1. The function includes two key adjustable parameters: a slope parameter that controls the steepness of the curve's ascent and a cutoff threshold that determines when the risk begins to rise significantly. These parameters are not empirically set but scientifically determined through optimization and fitting using maximum likelihood estimation on carefully collected historical training data from the system development phase. The training data must contain a sufficient number of historical patients undergoing cardiopulmonary bypass surgery with complete perioperative information (covering all pre-defined dimensions) and rigorously defined postoperative delirium syndrome outcome labels. To ensure optimal discriminative performance from the parameter combination, receiver operating characteristic (ROC) curves are plotted on independent validation set data to systematically evaluate the model performance under different parameter combinations. The final parameter combination that maximizes the area under the curve and achieves the highest Youden index is selected. This process maximizes the accuracy and robustness of the model in distinguishing between high-risk and low-risk patients with postoperative delirium.

[0023] Considering the evolution of clinical practice and the emergence of new evidence, the system features a dynamic update module to maintain its long-term predictive efficacy. This module receives real-time assessments of consciousness status based on the clinical gold standard CAM-ICU scale within 72 hours post-surgery. When the cumulative number of newly added, rigorously confirmed postoperative delirium events reaches a preset threshold, the system automatically triggers a model retraining process. The preset threshold is typically set at ≥5 cases, balancing the timeliness of model updates with the need for new data accumulation. During retraining, newly added samples containing complete data and outcome labels are added to the original training set, and the entire analysis process, from univariate analysis screening in the risk identification module to multivariate logistic regression modeling and variable selection in the risk assessment quantification module, is re-executed. Finally, based on the updated model results, the integer weights of each independent risk factor in the risk score generation module are recalculated and updated, and the slope parameter and cutoff threshold of the S-curve function in the risk probability calculation module are re-optimized using the new training dataset. This closed-loop, adaptive design allows the system to continuously learn new knowledge and dynamically adapt to changes in the clinical environment, thereby continuously improving the accuracy of predictions and its clinical applicability.

[0024] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A risk prediction system for postoperative delirium syndrome in cardiac surgery patients, characterized in that, include: The clinical data acquisition module is used to collect multidimensional clinical data of patients undergoing cardiac surgery under cardiopulmonary bypass during the perioperative period. The data includes patient baseline information, preoperative neurocognitive function assessment indicators, intraoperative physiological parameters, and early postoperative vital signs. The risk identification module is used to screen candidate risk factors that are significantly associated with delirium syndrome from the collected multidimensional clinical data using a univariate statistical analysis mechanism. The risk assessment quantification module is used to determine independent risk factors and their degree of influence by inputting candidate risk factors into a multivariate logistic regression model and based on the backward stepwise regression method. The risk score generation module is used to calculate the patient's individualized total risk score based on the independent risk factors identified in the assessment and their degree of influence. The risk probability calculation module is used to convert the total risk score into the incidence probability value of delirium syndrome.

2. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 1, characterized in that, The clinical data acquisition module collects multidimensional clinical data specifically as follows: The patient's baseline information includes, but is not limited to: age, gender, years of education, history of preoperative cognitive impairment, history of use of psychotropic drugs, and history of alcoholism; The preoperative neurocognitive function assessment indicators include, but are not limited to: Montreal Cognitive Assessment Scale score, Hamilton Anxiety Rating Scale score, and EEG alpha wave power percentage. The intraoperative physiological parameters include, but are not limited to: duration of cardiopulmonary bypass, maximum decrease in cerebral oxygen saturation from baseline, mean arterial pressure fluctuation coefficient, and cumulative remifentanil dosage. The early postoperative vital signs include, but are not limited to: Richmond agitation-sedation score, interleukin-6 serum concentration, and number of sleep-wake cycle changes within 24 hours.

3. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 1, characterized in that, The risk identification module uses the following method to screen candidate risk factors: We collected perioperative multidimensional clinical data from all patients undergoing cardiac surgery under cardiopulmonary bypass, preprocessed the multidimensional clinical data, and used the occurrence of delirium syndrome as a binary outcome variable. Perform independent statistical analysis for each clinical variable, and select the statistical test method according to the variable type: For continuous variables that conform to a normal distribution, the independent samples t-test is used; for continuous variables that do not conform to a normal distribution, the Mann-Whitney U test is used; the continuous variables include, but are not limited to: age, Montreal Cognitive Assessment Scale score, and the maximum decrease in brain oxygen saturation from baseline; Chi-square test was used for categorical variables; the categorical variables include, but are not limited to: gender, history of preoperative cognitive impairment, and history of alcoholism. Set a significance level threshold, calculate the p-value for each variable test. If the p-value is less than the significance level threshold, it indicates that the variable is statistically significantly associated with delirium syndrome and is included in the candidate risk factor set. All variables with significant p-values ​​were summarized as candidate risk factors and used as input for the subsequent multivariate logistic regression model.

4. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 1, characterized in that, The risk assessment quantification module determines independent risk factors and their degree of influence using the following method: All candidate risk factors identified through univariate analysis were incorporated into the initial multivariate logistic regression model, and the iterative calculation process was initiated: In each step, a likelihood ratio test is performed on all variables in the current model, and the variable with the largest p-value that is higher than the preset removal threshold is removed, with only one variable removed at a time. Repeat this elimination process until the p-values ​​of all remaining variables are less than the significance level threshold; After each removal, the model is refitted and the model fit is verified. The variables that are finally retained are the independent risk factors. For each identified independent risk factor, its adjusted odds ratio (OR) and confidence interval were calculated. The OR value represents the degree of influence of the factor on the risk of developing delirium syndrome, and the regression coefficient β value was recorded for subsequent scoring system construction.

5. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 4, characterized in that, The method for determining independent risk factors and their degree of influence requires collinearity diagnosis throughout the entire process. The specific method is as follows: After obtaining the final set of independent risk factors through backward stepwise regression, a linear regression model containing all identifiable factors is constructed. For each independent variable Xᵢ in the model, all other independent variables are used as predictors and Xᵢ is used as the dependent variable to establish an auxiliary linear regression model, and the determination coefficient R²ᵢ of the auxiliary model is calculated. Based on the calculated coefficient of determination, calculate the variance inflation factor of the independent variable; After completing the above calculations for all independent variables, check the VIF value of each variable: if all VIF values ​​are less than the preset value, it is determined that there is no significant multicollinearity; if there is a variable with VIF ≥ the preset value, identify the variable and assess its clinical significance, and delete the variable. After processing, rebuild the regression model and repeat the above VIF calculation process until the VIF values ​​of all variables in the final model are stable below the preset value; The final output consists of independent risk factors that pass the collinearity test and their regression coefficients.

6. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 1, characterized in that, The risk score generation module calculates the patient's individualized total risk score using the following method: Find the minimum value among all independent risk factor regression coefficients, calculate the weight value of each factor, which is equal to the regression coefficient of the factor divided by the minimum regression coefficient, and round the result to the nearest integer. The total risk score is obtained by summing the products of the weights of all independent risk factors and their indications of presence or absence, where the indications are determined according to the risk factor determination criteria preset by the system.

7. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 1, characterized in that, The method for converting the total risk score into the probability value of delirium syndrome is as follows: Receive the patient's individualized total risk score from the risk score generation module, and substitute the total risk score into the risk prediction function for calculation; The risk prediction function is an S-shaped curve function, and its output value represents the probability of disease occurrence; Key parameters in the function include: the slope parameter that controls the steepness of the curve and the cutoff threshold that determines the risk threshold. The values ​​of the key parameters are determined by applying the maximum likelihood estimation method to the training set data for optimization and fitting.

8. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 7, characterized in that, The slope parameter and cutoff threshold are determined by plotting the subject operating characteristic curve on the validation set and selecting the parameter combination that maximizes the area under the curve and has the highest Youden index.

9. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 7, characterized in that, The training set data specifically includes: Complete perioperative data of patients undergoing cardiopulmonary bypass surgery in the past were collected during the system development phase. The data included all dimensions, including: patient baseline information, preoperative neurocognitive assessment, intraoperative physiological parameters, and early postoperative vital signs. The key outcome label included whether the patient developed delirium syndrome postoperatively.

10. The risk prediction system for postoperative delirium syndrome in cardiac surgery patients according to claim 1, characterized in that, It also includes a dynamic update module, which connects the risk grading and intervention modules, specifically: Real-time reception of consciousness status assessment results based on the CAM-ICU scale within 72 hours postoperatively; When the cumulative number of newly added delirium events is ≥5, the model is retrained, the newly added data is added to the training set, and the single factor analysis of the risk identification module and the multiple regression of the risk assessment quantification module are re-executed. Update the weight values ​​of the risk scoring generation module and the slope parameters and cutoff threshold of the risk probability calculation module.

Images