Screening method, model and system for perioperative hypoglycemia prediction factors of digestive endoscope
By identifying seven predictive factors for hypoglycemia and constructing a logistic regression model, the standardization problem of perioperative hypoglycemia risk management in digestive endoscopy was solved, enabling rapid and simple hypoglycemia risk assessment, reducing the rate of missed diagnosis, and improving patient safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI EAST HOSPITAL EAST HOSPITAL TONGJI UNIV SCHOOL OF MEDICINE
- Filing Date
- 2026-05-07
- Publication Date
- 2026-06-02
AI Technical Summary
In the current technology, there is a lack of standardized prediction tools for the risk management of perioperative hypoglycemia in gastrointestinal endoscopy, and the existing models have poor universality, high rate of missed diagnosis, and are difficult to apply in the busy clinical work of nurses.
Seven predictive factors for hypoglycemia were identified using univariate analysis and multivariate logistic regression, including blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnostic category, and diastolic blood pressure. A multivariate logistic regression model was then constructed to assess the risk of hypoglycemia.
It provides a universally applicable hypoglycemia risk assessment system that can quickly and easily provide clinicians with hypoglycemia risk assessment, reduce the rate of missed diagnosis, and improve the safety of perioperative patients.
Smart Images
Figure CN122136005A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to medical technology, and in particular to a method, model, and system for screening predictive factors of perioperative hypoglycemia in digestive endoscopy. Background Technology
[0002] Endoscopic surgery (such as endoscopic mucosal resection and polypectomy) has become a common method for managing digestive system diseases. However, perioperative hypoglycemia is a common complication, with an incidence of 3%-18% in hospitalized patients, potentially leading to nerve damage, increased risk of infection, and prolonged hospital stay. Current techniques for managing hypoglycemia risk rely heavily on nurses' subjective experience, lacking standardized predictive tools. Although guidelines from the American Diabetes Association and other organizations emphasize personalized glycemic management, most models are designed for surgical patients and lack specificity for digestive endoscopy. Furthermore, traditional methods have a high rate of missed diagnoses (up to 30%) and an average intervention delay of 45 minutes, highlighting the urgent need to develop accurate predictive tools.
[0003] In recent years, machine learning technology has been used for risk prediction, but existing models often require complex indicators (such as more than 12 laboratory parameters), making them difficult to integrate into nurses' busy clinical work. For example, some models are only applicable to colonoscopy patients or diabetic patients, lacking general applicability. Therefore, there is an urgent need for a predictive system based on indicators that nurses can easily and quickly obtain and that are interpretable, in order to improve perioperative patient safety. Summary of the Invention
[0004] In view of the deficiencies in the existing technology, the technical problem to be solved by the present invention is to provide a method, model and system for screening predictive factors of perioperative hypoglycemia in digestive endoscopy that has good universality, can provide hypoglycemia risk assessment for perioperative patients and assist clinicians in decision-making.
[0005] To address the aforementioned technical problems, this invention provides a method for screening predictive factors of perioperative hypoglycemia in digestive endoscopy, characterized by the following specific steps:
[0006] Step 1: Select multiple factors that affect perioperative hypoglycemia during digestive endoscopy and define them as candidate factors;
[0007] We obtained sample data from multiple patients who underwent digestive endoscopy, including various candidate factors. With the occurrence of perioperative hypoglycemia as the outcome event, we used univariate analysis to analyze the sample data and screened out statistically significant candidate factors, which were defined as initial screening factors.
[0008] The method for determining whether a patient has perioperative hypoglycemia is as follows: if the patient's blood glucose level is below 3.9 mmol / L during the perioperative period, the patient is considered to have perioperative hypoglycemia; otherwise, the patient is considered not to have perioperative hypoglycemia. The perioperative period includes the 24 hours before the patient undergoes digestive endoscopy, the intraoperative period, and the 24 hours after the patient undergoes the procedure.
[0009] Step 2: Using whether the patient experiences perioperative hypoglycemia as the dependent variable and each initial screening factor as the independent variable, the sample data is analyzed using multivariate logistic regression. The initial screening factors with statistical significance are defined as hypoglycemia predictors.
[0010] Seven predictive factors for hypoglycemia were identified: blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnostic category, and diastolic blood pressure.
[0011] The present invention also provides a perioperative hypoglycemia risk prediction model for digestive endoscopy, characterized in that: the model is a multivariate logistic regression model, the outcome event of the model is set as whether the patient experiences perioperative hypoglycemia, and the input factors of the model are set as 7 hypoglycemia predictors.
[0012] The seven predictors of hypoglycemia were blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnosis category, and diastolic blood pressure.
[0013] The present invention also provides a perioperative hypoglycemia risk prediction system for digestive endoscopy, characterized in that it includes a data input module and a hypoglycemia prediction module;
[0014] The data input module is used to import data on seven hypoglycemia predictors for the subject. These seven hypoglycemia predictors include blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnosis category, and diastolic blood pressure.
[0015] The hypoglycemia prediction module has a built-in hypoglycemia risk prediction model. The hypoglycemia risk prediction model is used to analyze the hypoglycemia prediction factor data imported by the data input module, and output the prediction result of whether the subject will experience perioperative hypoglycemia based on the analysis results.
[0016] Furthermore, it also includes a nomogram construction module, which is used to construct a hypoglycemia risk nomogram containing weighted line segments of 7 hypoglycemia predictive factors based on the analysis results of the hypoglycemia risk prediction model, and output the constructed hypoglycemia risk nomogram to a display device for display.
[0017] The present invention provides a method, model, and system for screening predictive factors of perioperative hypoglycemia in digestive endoscopy. It uses univariate analysis combined with multivariate logistic regression to screen out factors that have an impact on the risk of perioperative hypoglycemia and incorporate them into the model to predict the risk of perioperative hypoglycemia. It can provide hypoglycemia risk assessment for perioperative patients, assist clinicians in decision-making, and the factors involved in the prediction are all routine indicators that nurses can easily and quickly obtain, with good universality and applicability to perioperative patients. Attached Figure Description
[0018] Figure 1 This is an example of an embodiment of the invention where sample data from the modeling group and the validation group are input into a hypoglycemia risk prediction model, and the resulting ROC curves are plotted.
[0019] Figure 2 This is a calibration curve plot drawn by inputting the sample data of each patient in the modeling group into the hypoglycemia risk prediction model in an embodiment of the present invention;
[0020] Figure 3 This is a calibration curve plot drawn by inputting the sample data of each patient in the validation group into the hypoglycemia risk prediction model in an embodiment of the present invention;
[0021] Figure 4 This is a decision curve (DCA) diagram of the hypoglycemia risk prediction model according to an embodiment of the present invention;
[0022] Figure 5 This is a clinical impact curve of the hypoglycemia risk prediction model according to an embodiment of the present invention;
[0023] Figure 6 This is a hypoglycemia risk prediction nomogram constructed from the analysis results of a patient sample data using a hypoglycemia risk prediction model according to an embodiment of the present invention.
[0024] Figure 7 This is a schematic diagram illustrating the influence of SHAP value features on the hypoglycemia risk prediction model in an embodiment of the present invention. Detailed Implementation
[0025] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings. However, these embodiments are not intended to limit the present invention. Any similar structures or variations thereof that adopt the present invention should be included in the protection scope of the present invention. The commas in the present invention all indicate the relationship between and. The English letters in the present invention are case-sensitive.
[0026] The method for screening predictive factors of perioperative hypoglycemia in gastrointestinal endoscopy provided by this invention includes the following specific steps:
[0027] Step 1: Select multiple factors that affect perioperative hypoglycemia during digestive endoscopy and define them as candidate factors;
[0028] We obtained sample data from multiple patients who underwent digestive endoscopy, including various candidate factors. With the occurrence of perioperative hypoglycemia as the outcome event, we used univariate analysis to analyze the sample data and screened out statistically significant candidate factors (p value less than 0.05) as the initial screening factors.
[0029] The method for determining whether a patient has perioperative hypoglycemia is as follows: if the patient's blood glucose level is below 3.9 mmol / L during the perioperative period, the patient is considered to have perioperative hypoglycemia; otherwise, the patient is considered not to have perioperative hypoglycemia. The perioperative period includes the 24 hours before the patient undergoes digestive endoscopy, the intraoperative period, and the 24 hours after the patient undergoes the procedure.
[0030] Step 2: Using whether the patient experiences perioperative hypoglycemia as the dependent variable and each initial screening factor as the independent variable, the sample data is analyzed using multivariate logistic regression. Statistically significant initial screening factors (p-value less than 0.05) are defined as hypoglycemia predictors.
[0031] Seven predictive factors for hypoglycemia were identified: blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnostic category, and diastolic blood pressure.
[0032] In this embodiment of the invention, by consulting clinical experts and combining literature reports, potential risk factors covering four major areas—demographics, vital signs, nutritional status, and clinical and metabolic indicators—were selected as candidate factors affecting perioperative hypoglycemia in digestive endoscopy.
[0033] Demographic factors include the patient's age and BMI (Body Mass Index). Age is divided into five categories: 18-44 years, 45-59 years, 60-74 years, 75-89 years, and ≥90 years. BMI is divided into three categories: underweight, normal weight, and overweight. BMI is calculated as weight (kg) ÷ height. 2 (m) 2 BMI < 18.5 is classified as underweight, BMI 18.5 ≤ BMI < 24 is classified as normal weight, and BMI ≥ 24 is classified as overweight.
[0034] Vital signs included the patient's systolic blood pressure, diastolic blood pressure, total protein, total bilirubin, alanine aminotransferase (ALT), total cholesterol, triglycerides, low-density lipoprotein (LDL-C), high-density lipoprotein (HDL-C), blood urea nitrogen (BUN), serum creatinine, glomerular filtration rate (eGFR), uric acid, preoperative blood glucose, and glycated hemoglobin (HbA1c). Systolic blood pressure was divided into four levels: 90-130 mmHg, 131-140 mmHg, 141-160 mmHg, and >160 mmHg. Diastolic blood pressure was divided into four levels: 60-90 mmHg, 91-100 mmHg, 101-110 mmHg, and >110 mmHg.
[0035] Nutritional status factors were assessed using the Nutritional Risk Screening 2002 (NRS-2002) score, a standardized assessment tool used in clinical nutrition diagnosis and treatment. A total score of ≥3 points was defined as having nutritional risk, and a total score of <3 points was defined as having no nutritional risk.
[0036] Clinical and metabolic factors include the patient's laxative use, diagnosis category, somatostatin use, type of endoscopic procedure, and history of comorbidities;
[0037] Laxative use status indicates whether the patient used laxatives during the perioperative period, and the value is "yes" or "no".
[0038] The diagnostic categories are divided into five categories: biliary and pancreatic diseases, rectal tumors, esophageal tumors, gastric tumors, and other diseases.
[0039] Somatostatin usage status indicates whether the patient has used somatostatin during the perioperative period, and the value is "yes" or "no".
[0040] Endoscopic procedures include six types: EFTR (endoscopic full-thickness resection), EMR (endoscopic mucosal resection), ERC (endoscopic retrograde cholangiopancreatography), ESD (endoscopic submucosal dissection), POEM (peroral endoscopic myotomy), and Others (other endoscopic procedures).
[0041] Comorbidity history is used to indicate whether a patient currently has two chronic diseases: diabetes and hypertension. There are four values: no chronic diseases (i.e., no diabetes and hypertension), only diabetes, only hypertension, and both chronic diseases (i.e., having both diabetes and hypertension).
[0042] In this embodiment of the invention, patients who underwent digestive endoscopy at Shanghai East Hospital from January 2022 to June 2024 were selected as volunteers to participate in the study and sample data including various candidate factors were obtained. The inclusion criteria for selected patients were: age ≥ 18 years and undergoing therapeutic digestive endoscopy. The exclusion criteria included: age < 18 years, incomplete clinical data, direct postoperative hospitalization, patients with hypoglycemia (blood glucose < 3.9 mmol / L) or critical illness.
[0043] A total of 3,534 patients were eventually included and randomly divided into two groups: the modeling group and the validation group. The modeling group included 2,476 patients, of whom 192 had hypoglycemia; the validation group included 1,060 patients, of whom 80 had hypoglycemia. There was no statistically significant difference in the incidence of hypoglycemia between the modeling group and the validation group (P>0.05).
[0044] All study participants were approved by the hospital's ethics review committee (approval number: 2024-046), complying with national ethics review standards, and the data were anonymized.
[0045] Using the occurrence of perioperative hypoglycemia as the outcome event, univariate analysis was used to analyze candidate factors in the sample data of each patient in the modeling group. Univariate analysis employed R software (version 4.3.1) and Python (version 3.10) for logistic regression. Rank-sum tests were used for continuous variables (such as blood glucose and BUN), and z-tests were used for categorical variables (such as laxative use and diagnostic category).
[0046] Table 1 shows the results of the univariate analysis of the modeling group. N in the table represents the number of people. The second column is the data of patients in the modeling group, the third column is the data of hypoglycemia-negative patients in the modeling group, and the fourth column is the data of hypoglycemia-positive patients in the modeling group.
[0047] The 13 variables, namely total protein, total bilirubin, alanine aminotransferase (ALT), total cholesterol, triglycerides, low-density lipoprotein (LDL-C), high-density lipoprotein (HDL-C), blood urea nitrogen (BUN), serum creatinine, glomerular filtration rate (eGFR), uric acid, preoperative blood glucose, and glycated hemoglobin (HbA1c), are continuous variables, while the other variables are categorical variables.
[0048] In each column of data for each continuous variable, the data outside the parentheses is the mean of all patients in that column, the left value inside the parentheses is the minimum value of all patients in that column, and the right value inside the parentheses is the maximum value of all patients in that column.
[0049] In each column of data for each categorical variable, the data outside the parentheses represents the number of patients, and the data inside the parentheses represents the percentage of patients with that variable in all patients in that column.
[0050] Table 1: Results of univariate analysis for the modeling group
[0051]
[0052]
[0053] As shown in Table 1, nine factors were statistically significant (P<0.05): blood urea nitrogen, glycated hemoglobin, glomerular filtration rate, diagnostic category, nutritional status (NRS-2002 score), diastolic blood pressure, serum creatinine, preoperative blood glucose, and laxative use. Therefore, these nine factors were defined as initial screening factors.
[0054] After screening statistically significant initial screening factors through univariate analysis, the initial screening factors of each patient sample data in the modeling group were analyzed using multivariate logistic regression with the patient's perioperative hypoglycemia as the dependent variable and each initial screening factor as the independent variable. Table 2 shows the results of multivariate logistic regression analysis of the modeling group.
[0055] Table 2: Results of Multivariate Logistic Regression Analysis for the Modeling Group
[0056]
[0057] As shown in Table 2, the seven initial screening factors (p value less than 0.05) – blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status (NRS-2002 score), laxative use, diagnostic category, and diastolic blood pressure – were statistically significant. Therefore, these seven initial screening factors were defined as hypoglycemic predictors.
[0058] This invention also provides a perioperative hypoglycemia risk prediction model for digestive endoscopy. This model is a multivariate logistic regression model. The outcome event of this model is set as whether the patient experiences perioperative hypoglycemia. The input factors of this model are set as the seven hypoglycemia predictors selected by the above method.
[0059] The values of seven hypoglycemia predictors from the patient samples in the modeling group were input into the hypoglycemia risk prediction model to plot ROC curves. The AUC value (area under the curve) was 0.907, and the confidence space (95% CI) was 0.883-0.931. The values of seven hypoglycemia predictors from the patient samples in the validation group were input into the hypoglycemia risk prediction model to plot ROC curves. The AUC value (area under the curve) was 0.872, and the confidence space (95% CI) was 0.826-0.917.
[0060] Figure 1The ROC curves are plotted by inputting sample data from the modeling group and validation group into the hypoglycemia risk prediction model. Figure 1 The blue curve in the image represents the ROC curve plotted using the sample data from the modeling group. Figure 1 The red curve in the figure is the ROC curve plotted by inputting the validation group sample data into the model;
[0061] from Figure 1 It can be seen that the model has excellent ability to distinguish the risk of hypoglycemia, which also reflects that the model has good generalization stability and can maintain reliable risk judgment effect in new object samples.
[0062] Figure 2 The calibration curve is plotted by inputting the values of seven hypoglycemia predictors from the patient sample data of each patient in the modeling group into the hypoglycemia risk prediction model. Figure 3 The calibration curve was plotted by inputting the values of seven hypoglycemia predictors from the patient sample data of the validation group into the hypoglycemia risk prediction model. Figure 2 , Figure 3 The horizontal axis represents the predicted probability of the outcome event by the hypoglycemia risk prediction model, the vertical axis represents the actual probability of the outcome event, the diagonal dashed line is the ideal diagonal line, representing that the predicted probability and the actual probability are always equal under ideal conditions, the blue curve is the original calibration line, and the red curve is the deviation correction line.
[0063] from Figure 2 , Figure 3 It can be seen that, based on the sample data of the modeling group and the validation group, the original calibration line has a high degree of fit with the ideal diagonal, proving that the predicted probability of the model has good consistency with the actual probability of occurrence.
[0064] Figure 4 This is a decision curve (DCA) plot of the hypoglycemia risk prediction model according to an embodiment of the present invention. It is used to characterize the clinically standardized net benefit of the hypoglycemia risk prediction model at different high-risk thresholds, so as to reflect the practical application value of the model. Figure 4 The horizontal axis represents the high-risk threshold, which is the critical value for determining the target object as hypoglycemic. The vertical axis represents the standardized net benefit, which is the quantitative ratio of the clinical benefit brought by correctly identifying hypoglycemic objects after the model is applied to the potential risk brought by incorrect identification. The higher the value, the stronger the clinical practical value of the model. Figure 4 The blue curve in the diagram represents the decision curve of the hypoglycemia risk prediction model in the modeling group sample, the pink curve represents the decision curve of the hypoglycemia risk prediction model in the validation group sample, the gray diagonal line represents the baseline of full intervention, and the black horizontal line represents the baseline of no intervention. The cost-benefit ratio axis on the lower side of the horizontal axis corresponds to the ratio of intervention cost to benefit value under various high-risk thresholds, which is used to match the decision-making needs of different clinical resource scenarios.
[0065] from Figure 4 It can be seen that the decision curves of the hypoglycemia risk prediction model in both the modeling group and the validation group are significantly higher than the baseline of full intervention and the baseline of no intervention, indicating that the application of the model can bring positive clinical net benefits in most high-risk threshold ranges. At the same time, combined with the cost-benefit ratio, it can be seen that the model can adapt to the decision-making needs of different resource investment scenarios, which reflects that the model has good clinical practical value.
[0066] Figure 5 This is a clinical impact curve of the hypoglycemia risk prediction model according to an embodiment of the present invention. It is used to characterize the number of hypoglycemic subjects identified and the actual number of hypoglycemic subjects under different high-risk thresholds, so as to reflect the efficiency of the model in identifying hypoglycemic individuals in clinical scenarios. Figure 5 The horizontal axis represents the high-risk threshold, and the vertical axis represents the number of hypoglycemic subjects (out of 1000), indicating the number of subjects identified as hypoglycemic by the model out of every 1000 target subjects. The blue curve represents the number of hypoglycemic subjects identified by the model at different high-risk thresholds, and the pink dashed line represents the actual number of hypoglycemic subjects at different high-risk thresholds. The cost-benefit ratio axis below the horizontal axis corresponds to the ratio of intervention cost to benefit value at various high-risk thresholds, used to match the decision-making needs of different clinical resource scenarios.
[0067] from Figure 5 It can be seen that as the high-risk threshold increases, the number of high-risk subjects identified by the model gradually decreases, but the proportion of subjects who actually experience hypoglycemia remains relatively stable. Combined with the cost-benefit ratio, it can be seen that the model can help clinicians select an appropriate high-risk threshold, achieving a balance between the efficiency of identifying high-risk subjects and the investment of clinical resources, demonstrating the operability and practicality of the model in actual clinical applications.
[0068] For each patient, the values of the patient's seven hypoglycemia predictors are input into the hypoglycemia risk prediction model for analysis. Based on the analysis results, a hypoglycemia risk prediction nomogram for the patient can be constructed and output to a visual device for easy viewing by medical staff. Figure 6 It is a hypoglycemia risk prediction nomogram constructed based on the analysis results of a patient sample data using a hypoglycemia risk prediction model. By comparing the weight segments of the seven hypoglycemia prediction factors in the nomogram with the scale markings on the score scale, the scores corresponding to the seven hypoglycemia prediction factors can be calculated. The total score is obtained by adding the scores of the seven hypoglycemia prediction factors together. The corresponding linear prediction value is obtained by mapping the total score to the linear prediction value, and the actual risk probability of hypoglycemia in the patient is obtained by mapping the linear prediction value to the linear prediction value.
[0069] Figure 7 This is a schematic diagram illustrating the influence of SHAP value features on the hypoglycemia risk prediction model according to an embodiment of the present invention. Figure 7 The horizontal axis represents the SHAP value, which is used to characterize the contribution of the corresponding feature (factor) to the prediction result of hypoglycemia risk. A positive SHAP value indicates that the feature will increase the hypoglycemia risk of the target object, and a negative SHAP value indicates that the feature will reduce the hypoglycemia risk of the target object. The color bar on the right is used to indicate the color corresponding to the high and low feature values. Yellow represents a high feature value and purple represents a low feature value.
[0070] Figure 7 The distribution range of SHAP values for the seven hypoglycemia predictors visually reflects the strength of these factors' contribution to the prediction of hypoglycemia risk. For example, when laxative use is at a high level (yellow), the SHAP value tends to be positive, indicating that laxative use increases the risk of hypoglycemia. Figure 7 It can be seen that the influence of the seven hypoglycemia predictors on the prediction results is consistent with clinical logic (such as the association between laxative use and hypoglycemia risk), while realizing the interpretability of the model prediction process and improving the credibility and practicality of the model in clinical applications.
[0071] This invention also provides a perioperative hypoglycemia risk prediction system for digestive endoscopy, characterized in that it includes:
[0072] The data input module is used to import data on seven hypoglycemia predictors for the test subjects. These seven hypoglycemia predictors include blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status (NRS-2002 score), laxative use, diagnosis category, and diastolic blood pressure.
[0073] The hypoglycemia prediction module has a built-in hypoglycemia risk prediction model constructed by the above method. The hypoglycemia risk prediction model is used to analyze the hypoglycemia prediction factor data imported by the data input module, and output the prediction result of whether the subject will experience perioperative hypoglycemia based on the analysis results.
[0074] The nomogram construction module is used to construct a hypoglycemia risk nomogram containing weighted line segments of 7 hypoglycemia predictive factors based on the analysis results of the hypoglycemia risk prediction model, and output the constructed hypoglycemia risk nomogram to a display device for display.
Claims
1. A method for screening predictive factors of perioperative hypoglycemia in digestive endoscopy, characterized in that, The specific steps are as follows: Step 1: Select multiple factors that affect perioperative hypoglycemia during digestive endoscopy and define them as candidate factors; We obtained sample data from multiple patients who underwent digestive endoscopy, including various candidate factors. With the occurrence of perioperative hypoglycemia as the outcome event, we used univariate analysis to analyze the sample data and screened out statistically significant candidate factors, which were defined as initial screening factors. The method for determining whether a patient has perioperative hypoglycemia is as follows: if the patient's blood glucose level is below 3.9 mmol / L during the perioperative period, the patient is considered to have perioperative hypoglycemia; otherwise, the patient is considered not to have perioperative hypoglycemia. The perioperative period includes the 24 hours before the patient undergoes digestive endoscopy, the intraoperative period, and the 24 hours after the patient undergoes the procedure. Step 2: Using whether the patient experiences perioperative hypoglycemia as the dependent variable and each initial screening factor as the independent variable, the sample data is analyzed using multivariate logistic regression. The initial screening factors with statistical significance are defined as hypoglycemia predictors. Seven predictive factors for hypoglycemia were identified: blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnostic category, and diastolic blood pressure.
2. A risk prediction model for perioperative hypoglycemia during digestive endoscopy, characterized in that: The model is a multivariate logistic regression model. The outcome event of the model is whether the patient experiences perioperative hypoglycemia. The input factors of the model are set as 7 hypoglycemia predictors. The seven predictors of hypoglycemia were blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnosis category, and diastolic blood pressure.
3. A perioperative hypoglycemia risk prediction system for digestive endoscopy, characterized in that: Includes a data input module and a hypoglycemia prediction module; The data input module is used to import data on seven hypoglycemia predictors for the subject. These seven hypoglycemia predictors include blood urea nitrogen, preoperative blood glucose, glycated hemoglobin, nutritional status, laxative use, diagnosis category, and diastolic blood pressure. The hypoglycemia prediction module has a built-in hypoglycemia risk prediction model of claim 2. The hypoglycemia risk prediction model is used to analyze the hypoglycemia prediction factor data imported by the data input module, and output the prediction result of whether the subject will experience perioperative hypoglycemia based on the analysis results.
4. The perioperative hypoglycemia risk prediction system for digestive endoscopy according to claim 3, characterized in that; It also includes a nomogram construction module, which is used to construct a hypoglycemia risk nomogram containing weighted line segments of 7 hypoglycemia predictive factors based on the analysis results of the hypoglycemia risk prediction model, and output the constructed hypoglycemia risk nomogram to a display device for display.