General anesthesia non-cardiac postoperative acute kidney injury score card model construction system and device and medium

Abstract

The invention discloses a system and a device for constructing an acute kidney injury score card model after a general anesthesia non-cardiac surgery and a medium, and the system comprises a database building unit which is used for building a disease database based on general anesthesia; the variable binning unit is used for extracting preoperative variables from the disease database, retaining original classifications of classified variables and carrying out quintile binning on continuous variables by adopting evidence weights; the variable screening unit is used for screening out preoperative variables with information values greater than 0.02; the variable dimensionality reduction unit is used for obtaining dimensionality-reduced variables by using the selected preoperative variables; the score card model establishing unit is used for establishing a score card model; and the risk classification unit is used for defining low-risk, medium-risk and high-risk patients by using different truncation values. According to the method, preoperative risk factors related to PO-AKI of patients are analyzed, the patients are divided into three levels of low risk, medium risk and high risk through different cut-off values, and then support is provided for subsequent clinical decision making.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent computer technology, and in particular relates to a system, device and medium for constructing a scoring card model for acute kidney injury after general anesthesia non-cardiac surgery. Background Art

[0002] Every year, more than 4 million people die within 30 days after surgery worldwide. Perioperative death has become the third leading cause of death in humans, second only to ischemic heart disease and stroke. Perioperative death is closely related to organ damage, among which acute kidney injury (AKI) is the most common. A large sample study of 3.6 million patients undergoing surgery showed that the incidence of postoperative acute kidney injury (PO-AKI) was 11.8%. Studies have shown that AKI may not only cause functional damage to other organs such as the heart and brain, but can also cause multiple organ failure and sepsis in severe cases. It is also closely related to prolonged hospitalization and increased short-term and long-term mortality. Severe AKI patients will develop chronic kidney disease and end-stage renal disease, which significantly increases the economic burden and affects the quality of life. At present, the clinical treatment of AKI is mainly symptomatic and supportive treatment. So far, there is still a lack of effective treatment to help patients restore kidney function to pre-injury levels. A meta-analysis pointed out that among patients with PO-AKI, about 26% of young patients and 31.3% of elderly patients did not have effective recovery of renal function. Therefore, it is particularly important to prevent the occurrence of PO-AKI. Summary of the invention

[0003] In order to solve the above technical problems, the present application provides a system, device and medium for constructing a scoring card model for postoperative acute kidney injury after general anesthesia non-cardiac surgery. By analyzing the preoperative risk factors related to PO-AKI of patients, the patients are divided into three levels: low risk, medium risk and high risk through different cutoff values, thereby providing support for subsequent clinical decision-making.

[0004] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present application.

[0005] According to a first aspect of the present invention, a scoring card model construction system for acute kidney injury after general anesthesia non-cardiac surgery is provided, the system comprising: A database building unit is configured to build a disease database based on clinical data of patients undergoing non-cardiac surgery under general anesthesia; a variable binning unit, configured to extract preoperative variables from the disease database, configured to extract preoperative variables from the disease database, the preoperative variables including categorical variables and continuous variables, retaining the original classification of the categorical variables, and binning the continuous variables into quintiles using weight of evidence (WOE); A variable screening unit is configured to calculate the information value (IV) of the preoperative variables after each quintile binning, and screen out preoperative variables with an information value greater than 0.02; The variable dimension reduction unit is configured to use the preoperative variables selected by the variable screening unit to train the least absolute shrinkage and selection operator (LASSO) model and perform a 10-fold cross-validation evaluation. The evaluation criteria include -2 times the natural logarithm of the likelihood function value and the area under the curve (AUC). The model with the highest AUC value within the range of 1 standard deviation is selected. , get the variable after dimension reduction; A scorecard model building unit is configured to build a Logistic model according to the variable after dimension reduction, and convert the Logistic model into a scorecard model; The risk classification unit is configured to define low-risk, medium-risk and high-risk patients using different cutoff values ​​based on the scorecard model.

[0006] Preferably, the variable binning unit is further configured to extract preoperative variables from the disease database, and perform quintile binning on the preoperative variables using weight of evidence using the following formula: ; In the formula, Indicates The corresponding bins value, Indicates The proportion of good samples in each bin, It indicates the proportion of good samples in the total sample. Indicates The proportion of bad samples in each bin, Indicates the proportion of bad samples in the total sample.

[0007] Preferably, the variable screening unit is further configured to calculate the information value of the preoperative variable after each quintile binning by the following formula: ; ; In the formula, Indicates IV value corresponding to each bin, Indicates Boxes, represents the information value of the feature, Indicates the total number of bins for this feature.

[0008] Preferably, the variable dimension reduction unit is further configured to obtain the variable after dimension reduction by the following formula: ; ; In the formula, It means to find the minimum solution of the objective function. represents the overall sample size, Indicates samples, represents the likelihood estimate of the sample point, Indicates Sample outcomes, Indicates feature variables, represents the characteristic coefficient vector, represents the estimated value of the model’s outcome indicator, represents the regularization coefficient, Expressed as a natural constant The exponential function with base , Represents the likelihood estimation function of the sample points.

[0009] Preferably, the scorecard model building unit is further configured to express the scorecard model as: ; In the formula, Indicates the scorecard score corresponding to a feature under a certain value. represents the benchmark fraction coefficient, represents the benchmark fraction coefficient, The positive sign in front makes the higher the probability of default, the higher the score, represents the probability of PO-AKI occurrence, Indicates the corresponding feature variable bin value, represents the coefficient of the variable in the model, represents the characteristic variable, Represents a constant.

[0010] Preferably, the risk classification unit is further configured to use the risk score as the cutoff value corresponding to the model sensitivity and specificity; select the risk score corresponding to the specificity greater than or equal to 90% as the critical value for low risk and medium risk; select the risk score corresponding to the sensitivity greater than or equal to 90% as the critical value for medium risk and high risk.

[0011] According to a second aspect of the present invention, a device for constructing a scoring card model for acute kidney injury after general anesthesia non-cardiac surgery is provided, comprising: Memory for storing computer programs; A processor is used to execute the computer program to implement the following steps: Establish a disease database based on the clinical data of patients undergoing non-cardiac surgery under general anesthesia; Extracting preoperative variables from the disease database, the preoperative variables include categorical variables and continuous variables, retaining the original classification of the categorical variables, and binning the continuous variables into quintiles using weight of evidence; The information value of the preoperative variables after each quintile bin was calculated, and preoperative variables with information value greater than 0.02 were screened; The preoperative variables screened by the variable screening unit were used to train the minimum absolute value compression and selection operator model and perform a 10-fold cross-validation evaluation. The evaluation criteria included -2 ​​times the natural logarithm of the likelihood function value and the area under the curve. The model with the highest AUC value within the range of 1 standard deviation was selected. , get the variable after dimension reduction; Establishing a Logistic model based on the variable after dimension reduction, and converting the Logistic model into a scorecard model; Based on the scorecard model, different cutoff values ​​were used to define low-risk, intermediate-risk, and high-risk patients.

[0012] According to a third aspect of the present invention, a readable storage medium is provided, wherein the readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the following steps: Establish a disease database based on the clinical data of patients undergoing non-cardiac surgery under general anesthesia; Extracting preoperative variables from the disease database, the preoperative variables include categorical variables and continuous variables, retaining the original classification of the categorical variables, and binning the continuous variables into quintiles using weight of evidence; The information value of the preoperative variables after each quintile bin was calculated, and preoperative variables with information value greater than 0.02 were screened; The preoperative variables screened by the variable screening unit were used to train the minimum absolute value compression and selection operator model and perform a 10-fold cross-validation evaluation. The evaluation criteria included -2 ​​times the natural logarithm of the likelihood function value and the area under the curve. The lambda corresponding to the highest AUC value within the range of 1 times the standard deviation was selected to obtain the variable after dimension reduction. Establishing a Logistic model based on the variable after dimension reduction, and converting the Logistic model into a scorecard model; Based on the scorecard model, different cutoff values ​​were used to define low-risk, intermediate-risk, and high-risk patients.

[0013] Beneficial effects: The present invention discloses a scoring card model construction system for acute kidney injury after general anesthesia non-cardiac surgery. By comprehensively analyzing multiple preoperative variables, IV value and Lasso regression are used to screen out important risk factors, helping doctors to more accurately identify high-risk patients. The scoring card model provides clinicians with an intuitive risk assessment tool. The model has a simple structure and a clear scoring calculation process. Clinical medical staff can quickly get started, which is convenient for promotion and application in actual operations. It is conducive to providing patients with personalized intervention suggestions before, during and after surgery, thereby optimizing clinical decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a structural diagram of a scoring card model construction system for acute kidney injury after general anesthesia non-cardiac surgery shown in an embodiment of the present invention; Figure 2 : is a graph showing the relationship between the LASSO regression partial likelihood function deviation and Log (λ) of the PO-AKI score card model shown in an embodiment of the present invention; Figure 3 It is a relationship diagram between the area under the LASSO curve and Log (λ) of the PO-AKI scoring card model shown in an embodiment of the present invention; Figure 4 PO-AKI score card model receiver operating characteristic curve diagram shown in the embodiment of the present invention; Figure 5 Schematic diagram of a PO-AKI scoring card shown in an embodiment of the present invention; in the figure: Elective / emergency, 1. Elective surgery, 2. Emergency surgery; Gender, 1: male, 2: female; Diabetes, 0: no history of diabetes, 1: non-insulin-dependent, 2: insulin-dependent; Exercise equivalent, 1: >6 metabolic equivalents, 2: 3-6 metabolic equivalents, 3: <3 metabolic equivalents; Hypertension, 0: no history of hypertension, 1: grade I, 2: grade II, 3: grade III; History of ischemic heart disease: 0: no chest pain or tightness after activity in the past 3 months; 1: history of non-myocardial infarction chest pain in the past 3 months; 2: acute myocardial infarction > 3 months; 3: acute myocardial infarction < 3 months; General condition, 1: good, 2: average, 3: poor; Consciousness state, 1: awake, 2: drowsiness, 3: stupor, 4: light coma, 5: deep coma; Surgical site, 1: limb orthopedic surgery, 2: spinal surgery, 3: body surface surgery, 4: cranial surgery, 5: head, face and neck surgery, 6: joint orthopedic surgery, 7: lower abdominal surgery, 8: thoracic surgery, 9: major vascular surgery, 10: upper abdominal surgery, 11: combined thoracoabdominal surgery, 12: total abdominal surgery; American Society of Anesthesiologists (ASA) classification: 1: I, 2: II, 3: III, 4: IV, 5: V; Figure 6 It is a structural diagram of a scoring card model construction device for acute kidney injury after general anesthesia non-cardiac surgery shown in an embodiment of the present invention; Figure 7 This is a flow chart of constructing a scoring card model for acute kidney injury after non-cardiac surgery under general anesthesia, as shown in an embodiment of the present invention.

[0015] Description of the drawings: 100 - system, 101 - database establishment unit, 102 - variable binning unit, 103 - variable screening unit, 104 - variable dimension reduction unit, 105 - scoring card model establishment unit, 106 - risk classification unit; 601 - memory, 602 - processor. DETAILED DESCRIPTION

[0016] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. The technical problems solved by the technical solution of the present invention and the beneficial effects are also described. It should be noted that the described embodiments are only intended to facilitate the understanding of the present invention and do not have any limiting effect on it.

[0017] At present, in the prior art, the Chinese patent publication number CN117153380A discloses a prediction method, system and device for acute kidney injury after non-cardiac surgery. The system first divides patients into high-risk patients and non-high-risk patients based on clinical experience, and then predicts the risk of acute kidney injury after surgery respectively, providing certain support for clinical decision-making. However, the system has some significant disadvantages.

[0018] First, the system distinguishes high-risk patients from non-high-risk patients through clinical experience. The clinical information selected for prediction may not be accurate enough and is not suitable for all study populations. Therefore, a unified variable screening method is needed to apply to different study populations. Second, the internal working mechanism of the high-risk patient prediction model is complex, making it difficult for clinicians to make independent risk judgments for postoperative acute kidney injury without the system.

[0019] Based on this, an embodiment of the present invention provides a system, device and medium for constructing a scoring card model for acute kidney injury after postoperative non-cardiac surgery under general anesthesia to solve the defects of the prior art in variable screening, interpretability and applicability. The system uses information value to perform preliminary screening of candidate model variables, and then reduces the dimension of the variables by minimum absolute value compression and selection operator. A logistic regression model is established based on the reduced variables and converted into a standard scoring card format. The model constructed by the present invention allows doctors to calculate the score by themselves without relying on complex systems. The goal of the present invention is to improve the applicability of the PO-AKI prediction model, thereby improving the clinical outcomes of patients and the efficiency of medical resource utilization.

[0020] Figure 1 This is a structural diagram of the system for constructing a scoring card model for acute kidney injury after surgery under general anesthesia and non-cardiac surgery in this embodiment; an embodiment of the present invention provides a scoring card model construction system for acute kidney injury after surgery under general anesthesia and non-cardiac surgery, the system 100 includes a database establishment unit 101, a variable binning unit 102, a variable screening unit 103, a variable dimension reduction unit 104, a scoring card model establishment unit 105 and a risk classification unit 106.

[0021] The database establishing unit 101 is configured to establish a disease database based on clinical data of patients undergoing non-cardiac surgery under general anesthesia.

[0022] The variable binning unit 102 is configured to extract preoperative variables from the disease database, wherein the preoperative variables include categorical variables and continuous variables, retain the original classification of the categorical variables, and bin the continuous variables into quintiles using evidence weights.

[0023] In this embodiment, after extracting the preoperative variables, the variable binning unit 102 retains the original classification for categorical variables and uses the weight of evidence to perform quintile binning for continuous variables. The formula is as follows: ; The variable screening unit 103 is configured to calculate the information value of the preoperative variables after each quintile binning, and screen out the preoperative variables with an information value greater than 0.02.

[0024] In some embodiments, the variable screening unit 103 is further configured to calculate the information value of the preoperative variable after each quintile binning by the following formula: ; ; In the formula, Indicates IV value corresponding to each bin, Indicates Boxes, represents the information value of the feature, Indicates the total number of bins for this feature.

[0025] In some embodiments, the variable screening unit 103 uses the Kolmogorov-Smirnov test to evaluate the data normality of the preoperative variables. Normally distributed measurement data are expressed as mean ± standard deviation, and non-normally distributed measurement data are expressed as median and interquartile range.

[0026] The variable dimension reduction unit 104 is configured to use the preoperative variables screened by the variable screening unit to train the minimum absolute value compression and selection operator model and perform a 10-fold cross validation evaluation. The evaluation criteria include -2 times the natural logarithm of the likelihood function value and the area under the curve. The λ corresponding to the highest AUC value within the range of 1 times the standard deviation is selected to obtain the reduced variable.

[0027] In some embodiments, the variable dimension reduction unit 104 is further configured to obtain the variable after dimension reduction by the following formula: ; ; In the formula, It means to find the minimum solution of the objective function. represents the overall sample size, Indicates samples, represents the likelihood estimate of the sample point, Indicates Sample outcomes, Indicates feature variables, represents the characteristic coefficient vector, represents the estimated value of the model’s outcome indicator, represents the regularization coefficient, Expressed as a natural constant The exponential function with base , Represents the likelihood estimation function of the sample points.

[0028] The scorecard model building unit 105 is configured to build a Logistic model according to the variable after dimension reduction, and convert the Logistic model into a scorecard model.

[0029] In some embodiments, the scorecard model building unit 105 is further configured to represent the scorecard model as: ; In the formula, Indicates the scorecard score corresponding to a feature under a certain value. represents the benchmark fraction coefficient, represents the benchmark fraction coefficient, The positive sign in front makes the higher the probability of default, the higher the score, represents the probability of PO-AKI occurrence, Indicates the corresponding feature variable bin value, represents the coefficient of the variable in the model, represents the characteristic variable, Represents a constant.

[0030] The risk classification unit 106 is configured to define low-risk, medium-risk and high-risk patients using different cutoff values ​​based on the scorecard model.

[0031] In some embodiments, the risk classification unit 106 is further configured to use the risk score as the cutoff value corresponding to the model sensitivity and specificity; select the risk score corresponding to the specificity greater than or equal to 90% as the critical value for low risk and medium risk; select the risk score corresponding to the sensitivity greater than or equal to 90% as the critical value for medium risk and high risk.

[0032] The feasibility and progress of the present invention will be demonstrated by a specific implementation case.

[0033] A total of 34,668 adult patients who underwent non-cardiac surgery under general anesthesia in West China Hospital of Sichuan University from June 25, 2019 to December 31, 2021 were sorted and collected, and a total of 131 preoperative variables of the patients were collected and sorted.

[0034] The above-mentioned preoperative variables are processed based on the scoring card model construction system for acute kidney injury after non-cardiac surgery under general anesthesia provided in the above embodiment.

[0035] Among them, acute kidney injury is defined as an increase in creatinine by ≥0.3 mg / dl (≥26.5 μmol / L) within 48 hours after surgery compared with the preoperative creatinine or an increase in creatinine by ≥1.5 times the preoperative creatinine within 7 days after surgery.

[0036] After completing the WOE conversion of the variables, the IV values ​​of all variables were calculated, and a total of 104 variables with IV values ​​greater than 0.02 were screened out.

[0037] The variables obtained by the preliminary screening of IV values ​​were subjected to LASSO regression. After 10-fold cross validation, the parameter adjustment relationship of LASSO regression was obtained as follows: Figure 2 and Figure 3 As shown in the figure, the λ corresponding to the maximum AUC value within the range of 1 standard deviation was selected, and the optimal tolerable Log (λ) value was -6.063649, and a total of 27 candidate variables were obtained after LASSO regression dimensionality reduction.

[0038] The 27 variables after dimensionality reduction included American Society of Anesthesiologists grade, elective / emergency, surgical site, serum cystatin C, history of ischemic heart disease, creatinine, diastolic blood pressure, consciousness status, estimated glomerular filtration rate, general condition, age, albumin, coefficient of variation of red blood cell distribution width, exercise equivalent, glutamyl transpeptidase, systolic blood pressure, heart rate, absolute lymphocyte count, red blood cell count, diabetes, prothrombin time, magnesium, white blood cell ratio, standard deviation of red blood cell distribution width, aspartate aminotransferase, hypertension and gender.

[0039] Based on the 27 variables screened, the PO-AKI score card model was established by logistic regression, and the model showed good predictive effect, with an average AUROC of 0.80 (95% CI: 0.78-0.82) ( Figure 4 To facilitate clinical practice, all variables included in the model were used to calculate the score card scores and compiled into a PO-AKI score card, see Figure 5 .

[0040] According to the cutoff values ​​corresponding to the high sensitivity (90%) and high specificity (90%) of the model, the patients were divided into three strata: low risk of PO-AKI, <15 points; intermediate risk of PO-AKI, 15-67 points; and high risk of PO-AKI, >67 points.

[0041] The number of low-risk patients was the largest, totaling 16,586, and their PO-AKI incidence was also the lowest at 0.49%; the number of medium-risk patients was 14,380, and their PO-AKI incidence was 2.48%, which was closest to the PO-AKI incidence of the entire population (2.43%); the number of high-risk patients was the smallest, totaling 3,702, and their PO-AKI incidence was also the highest at 10.99%.

[0042] The embodiment of the present invention also provides a scoring card model construction device for acute kidney injury after general anesthesia non-cardiac surgery, such as Figure 6 As shown, the device includes: a memory 601 for storing a computer program; a processor 602 for executing the computer program to implement the following Figure 7The steps for constructing the score card model for acute kidney injury after non-cardiac surgery under general anesthesia are shown: S1. Establish a disease database based on the clinical data of patients undergoing non-cardiac surgery under general anesthesia; S2. extracting preoperative variables from the disease database, wherein the preoperative variables include categorical variables and continuous variables, retaining the original classification of the categorical variables, and binning the continuous variables into quintiles using weight of evidence; S3. Calculate the information value of the preoperative variables after each quintile binning, and select the preoperative variables with information value greater than 0.02; S4. Use the preoperative variables screened by the variable screening unit to train the minimum absolute value compression and selection operator model and perform a 10-fold cross validation evaluation. The evaluation criteria include -2 times the natural logarithm of the likelihood function value and the area under the curve. Select the λ corresponding to the highest AUC value within the range of 1 times the standard deviation to obtain the variable after dimension reduction. S5. Establish a Logistic model based on the variable after dimension reduction, and convert the Logistic model into a scorecard model; S6. Based on the scorecard model, different cutoff values ​​are used to define low-risk, medium-risk and high-risk patients.

[0043] The embodiment of the present invention also provides a readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the following steps: establishing a disease database based on the clinical data of patients undergoing non-cardiac surgery under general anesthesia; extracting preoperative variables from the disease database, the preoperative variables including categorical variables and continuous variables, retaining their original classifications for the categorical variables, and using evidence weights to perform quintile binning for the continuous variables; calculating the information value of the preoperative variables after each quintile binning, and screening out preoperative variables with information values ​​greater than 0.02; using the preoperative variables screened out by the variable screening unit to train the minimum absolute value compression and selection operator model and perform a 10-fold cross-validation evaluation, the evaluation criteria including -2 times the natural logarithm of the likelihood function value and the area under the curve, selecting the λ corresponding to the highest AUC value within the range of 1 times the standard deviation, and obtaining the variable after dimension reduction; establishing a Logistic model based on the variable after dimension reduction, and converting the Logistic model into a scorecard model; based on the scorecard model, using different cutoff values ​​to define low-risk, medium-risk and high-risk patients The specific description above further illustrates the purpose, technical solutions and beneficial effects of the invention in detail. It should be understood that the above is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A scoring card model construction system for acute kidney injury after general anesthesia non-cardiac surgery, characterized in that: The system comprises: A database building unit is configured to build a disease database based on clinical data of patients undergoing non-cardiac surgery under general anesthesia; A variable binning unit is configured to extract preoperative variables from the disease database, wherein the preoperative variables include categorical variables and continuous variables, retain the original classification of the categorical variables, and perform quintile binning on the continuous variables using weight of evidence; A variable screening unit is configured to calculate the information value of the preoperative variables after each quintile binning, and screen out the preoperative variables with an information value greater than 0.02; The variable dimension reduction unit is configured to use the preoperative variables screened by the variable screening unit to train the minimum absolute value compression and selection operator model and perform a 10-fold cross validation evaluation. The evaluation criteria include -2 times the natural logarithm of the likelihood function value and the area under the curve. The λ corresponding to the highest AUC value within the range of 1 times the standard deviation is selected to obtain the variable after dimension reduction. Among them, AUC is an indicator for evaluating the effect of the model; λ represents the Lasso regularization coefficient. A scorecard model building unit is configured to build a Logistic model according to the variable after dimension reduction, and convert the Logistic model into a scorecard model; The risk classification unit is configured to define low-risk, medium-risk and high-risk patients using different cutoff values ​​based on the scorecard model.

2. The system according to claim 1, characterized in that The variable binning unit is further configured to extract preoperative variables from the disease database, and bin the preoperative variables into quintiles using the weight of evidence using the following formula: ； In the formula, Indicates The corresponding bins value, Indicates The proportion of good samples in each bin, It indicates the proportion of good samples in the total sample. Indicates The proportion of bad samples in each bin, Indicates the proportion of bad samples in the total sample.

3. The system according to claim 2, characterized in that The variable screening unit is further configured to calculate the information value of the preoperative variable after each quintile binning by the following formula: ； ； In the formula, Indicates IV value corresponding to each bin, Indicates Boxes, represents the information value of the feature, Indicates the total number of bins for this feature.

4. The system according to claim 1, characterized in that The variable dimension reduction unit is further configured to obtain the variable after dimension reduction by the following formula: ； ； In the formula, It means to find the minimum solution of the objective function. represents the overall sample size, Indicates samples, represents the likelihood estimate of the sample point, Indicates Sample outcomes, Indicates feature variables, represents the characteristic coefficient vector, represents the estimated value of the model’s outcome indicator, represents the regularization coefficient, Expressed as a natural constant The exponential function with base , Represents the likelihood estimation function of the sample points.

5. The system according to claim 1, characterized in that The scorecard model building unit is further configured to represent the scorecard model as: ； In the formula, Indicates the scorecard score corresponding to a feature under a certain value. represents the benchmark fraction coefficient, represents the benchmark fraction coefficient, The positive sign in front makes the higher the probability of default, the higher the score, represents the probability of PO-AKI occurrence, Indicates the corresponding feature variable bin value, represents the coefficient of the variable in the model, represents the characteristic variable, Represents a constant.

6. The system according to claim 1, characterized in that The risk classification unit is further configured to use the risk score as the cutoff value corresponding to the model sensitivity and specificity; select the risk score corresponding to the specificity greater than or equal to 90% as the critical value of low risk and medium risk; The risk score corresponding to a sensitivity greater than or equal to 90% was selected as the cutoff value for medium risk and high risk.

7. A scoring card model construction device for acute kidney injury after general anesthesia non-cardiac surgery, characterized in that: include: Memory for storing computer programs; A processor is used to execute the computer program to implement the following steps: Establish a disease database based on the clinical data of patients undergoing non-cardiac surgery under general anesthesia; Extracting preoperative variables from the disease database, the preoperative variables include categorical variables and continuous variables, retaining the original classification of the categorical variables, and binning the continuous variables into quintiles using weight of evidence; The information value of the preoperative variables after each quintile bin was calculated, and preoperative variables with information value greater than 0.02 were screened; The preoperative variables screened by the variable screening unit were used to train the minimum absolute value compression and selection operator model and perform a 10-fold cross-validation evaluation. The evaluation criteria included -2 ​​times the natural logarithm of the likelihood function value and the area under the curve. The lambda corresponding to the highest AUC value within the range of 1 times the standard deviation was selected to obtain the variable after dimension reduction. Establishing a Logistic model based on the variable after dimension reduction, and converting the Logistic model into a scorecard model; Based on the scorecard model, different cutoff values ​​were used to define low-risk, intermediate-risk, and high-risk patients.

8. A readable storage medium, characterized in that: The readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the following steps: Establish a disease database based on the clinical data of patients undergoing non-cardiac surgery under general anesthesia; Extracting preoperative variables from the disease database, the preoperative variables include categorical variables and continuous variables, retaining the original classification of the categorical variables, and binning the continuous variables into quintiles using weight of evidence; The information value of the preoperative variables after each quintile bin was calculated, and preoperative variables with information value greater than 0.02 were screened; The preoperative variables screened by the variable screening unit were used to train the minimum absolute value compression and selection operator model and perform a 10-fold cross-validation evaluation. The evaluation criteria included -2 ​​times the natural logarithm of the likelihood function value and the area under the curve. The model with the highest AUC value within the range of 1 standard deviation was selected. , get the variable after dimension reduction; Establishing a Logistic model based on the variable after dimension reduction, and converting the Logistic model into a scorecard model; Based on the scorecard model, different cutoff values ​​were used to define low-risk, intermediate-risk, and high-risk patients.

Images