Elderly patient postoperative serious complication risk probability prediction column graph model construction method based on preoperative human N-terminal brain natriuretic peptide

Abstract

The invention relates to the field of biomedicine, in particular to a construction method of an old patient postoperative serious complication risk probability prediction column graph model based on preoperative human N-terminal brain natriuretic peptide. According to the method, five independent risk factors related to the occurrence of the postoperative serious complications of the elderly patient are provided, and the risk probability prediction column graph model is constructed by using multi-factor logistic regression and is used for evaluating the risk of the postoperative serious complications of the elderly patient. The method comprises the following steps: firstly, collecting clinical data of an operation of an elderly patient; then risk factor screening is carried out through multi-single-factor logistic regression; then, ranking the risk factors according to importance by using an optimal subset regression method to obtain characteristic variables with relatively high prediction values; and finally, establishing a multi-factor logistic regression risk probability prediction model, and constructing a column graph according to the prediction model. According to the method, the occurrence risk of the postoperative serious complications of the elderly patient can be accurately predicted based on the pre-operative N-terminal brain natriuretic peptide, clinical doctors can be helped to recognize the high-risk elderly patient suffering from the postoperative serious complications in the early stage, the doctors are guided to conduct individualized intervention, and occurrence of the postoperative serious complications of the elderly patient is reduced.

Description

Technical Field

[0001] This invention relates to the field of biomedical technology, specifically to a method for constructing a nomogram model for predicting the probability of severe postoperative complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide. Background Technology

[0002] With an aging population, more and more elderly patients are undergoing surgery and anesthesia. Elderly patients often have comorbidities affecting various systems, including the heart, kidneys, and brain, resulting in a significantly higher risk of postoperative complications compared to adults of other ages, especially severe postoperative complications (Clavien-Dindo grade ≥3). Severe postoperative complications increase mortality, hospital stay, and hospitalization costs, reducing quality of life. Existing tools for predicting postoperative adverse event risks rely heavily on physicians' skills, experience, and subjective judgment, such as the American Society of Anesthesiologists (ASSA) classification and cardiac function index classification, lacking objective, simple, and intuitive risk stratification tools. Therefore, there is an urgent need for an early, rapid, objective tool that reflects the impact of surgery and anesthesia on postoperative complication risk assessment, enabling early warning and guiding perioperative clinical decision-making.

[0003] Human N-terminal brain natriuretic peptide (HNB) is a member of the natriuretic peptide family synthesized and secreted by cardiomyocytes in response to stimuli such as volume expansion and pressure overload. In recent years, HNB-targeted therapy and its predictive value for major postoperative cardiovascular adverse events and postoperative pulmonary complications have received considerable attention. Perioperative monitoring of plasma HNB concentrations can not only assess the patient's preoperative circulatory and cardiac function status but also predict the occurrence of postoperative complications, guiding clinicians to intervene early and prevent serious complications. Studies have shown that high preoperative HNB levels are a risk factor for postoperative complications in thoracic surgery. Furthermore, if HNB levels remain elevated within 5 days postoperatively, the risk of early postoperative complications (Clavien-Dindo grade ≥2 occurring during hospitalization) and late postoperative complications (Clavien-Dindo grade ≥2 occurring within 3 months postoperatively) significantly increases. Studies have found that serum levels of human N-terminal brain natriuretic peptide (HNT) are correlated with the amount of fluid therapy in patients undergoing colorectal resection; increased fluid volume leads to higher HNT concentrations. Furthermore, preoperative HNT levels can predict postoperative cardiopulmonary complications such as atrial fibrillation, ventricular fibrillation, respiratory failure, and pulmonary edema. Therefore, preoperative HNT is an effective biomarker for predicting postoperative complications in elderly patients. Currently, there is no model for predicting severe postoperative complications in elderly patients based on preoperative HNT. There is an urgent need to develop a probabilistic prediction model for the risk of severe postoperative complications in elderly patients based on preoperative HNT, one that can be easily used by clinicians to achieve early warning and intervention, improve postoperative outcomes in elderly patients, and reduce medical costs. Summary of the Invention

[0004] To address the shortcomings in the aforementioned background technology, this invention proposes a nomogram model for predicting the probability of postoperative serious complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide (BNP), thus solving the current problem of the lack of a nomogram model for predicting the probability of postoperative serious complications in elderly patients based on preoperative human BNP.

[0005] The technical solution of this invention is implemented as follows:

[0006] 1. A method for constructing a nomogram model for predicting the probability of postoperative serious complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide, the steps of which are as follows:

[0007] Step 1: Collect perioperative clinical data and follow-up data of elderly patients.

[0008] Step 2: Use univariate logistic regression to perform variable significance analysis on the aforementioned perioperative clinical data to screen out potential characteristic variables.

[0009] Step 3: Calculate the importance of the latent feature variables selected in Step 2 using optimal subset regression, and rank the latent feature variables selected in Step 2 according to their importance.

[0010] Step 4: Based on the reordered latent feature variables from Step 3, construct risk probability prediction models for different variable combinations using multifactor logistic regression, and calculate the AIC value of the risk probability prediction models for different variable combinations.

[0011] Step 5: Compare the risk probability prediction models of different variable combinations in Step 4 using the AIC value, and select the variable combination model with the smallest AIC value as the optimal risk probability prediction model.

[0012] Step Six: Use the ROC curve method to evaluate the accuracy of the optimal risk probability prediction model described in Step Five.

[0013] Step 7: Use the nomogram method to draw the nomogram model of the optimal risk probability prediction model described in Step 5.

[0014] The perioperative clinical data of the elderly patients included gender, age, weight, height, surgical site, operation duration, anesthesia duration, smoking history, alcohol consumption history, types and dosages of intraoperative medications, total intraoperative fluid infusion, intraoperative blood transfusion, preoperative hypertension, preoperative diabetes, American College of Anesthesiologists classification, exercise tolerance, breath-holding test classification, urine output, total length of hospital stay, length of intensive care unit stay, and preoperative human N-terminal brain natriuretic peptide concentration; the follow-up data included whether serious complications occurred within 30 days after surgery.

[0015] The potential characteristic variables selected include age, weight, duration of anesthesia, smoking history, total intraoperative fluid infusion, intraoperative blood transfusion, preoperative hypertension, preoperative diabetes, American College of Anesthesiologists classification, exercise tolerance, breath-holding test classification, total length of hospital stay, length of intensive care unit stay, and preoperative human N-terminal brain natriuretic peptide concentration.

[0016] The potential characteristic variables selected were sorted according to their importance in the following order: preoperative human N-terminal brain natriuretic peptide concentration, age, American College of Anesthesiologists classification, duration of anesthesia, weight, intraoperative blood transfusion volume, preoperative hypertension, preoperative diabetes, exercise tolerance, total length of hospital stay, length of intensive care unit stay, smoking history, total intraoperative fluid infusion volume, and breath-holding test classification.

[0017] The optimal risk probability prediction model includes the following characteristic variables: preoperative N-terminal brain natriuretic peptide concentration, age, American College of Anesthesiologists classification, duration of anesthesia, and weight.

[0018] The beneficial effects of this technical solution are:

[0019] (1) This invention uses univariate logistic regression analysis to screen potential characteristic variables from perioperative clinical data and follow-up data, and uses the optimal subset regression method to rank the importance of potential characteristic variables, thereby finding the most important characteristic variables for serious postoperative complications in elderly patients.

[0020] (2) Based on the selected characteristic variables, the present invention constructs a multi-factor logistic regression risk probability prediction model with different combinations of variables, uses the AIC value to determine the optimal risk probability prediction model, and can obtain the risk of serious postoperative complications in elderly patients through the optimal risk probability prediction model, enabling early warning and accurate identification of high-risk elderly patients. The risk probability prediction nomogram model is drawn by the nomogram method, which can be convenient for clinicians to use, intervene early, improve the postoperative outcome of elderly patients, and reduce medical costs. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of a risk probability prediction nomogram model provided in an embodiment of the present invention.

[0024] Figure 2 This is the overall flowchart of the present invention.

[0025] Figure 3 The ROC curve of the optimal risk probability prediction model established by the present invention is provided for embodiments of the present invention. Detailed Implementation

[0027] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0028] like Figure 1 As shown in the figure, this invention provides a method for constructing a nomogram model for predicting the probability of postoperative serious complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide. The specific steps are as follows:

[0029] Step 1: Collect perioperative clinical data and follow-up data of elderly patients; the perioperative clinical data of elderly patients includes gender, age, weight, height, surgical site, operation time, anesthesia time, smoking history, alcohol consumption history, total intraoperative fluid infusion, intraoperative blood transfusion, whether hypertension or diabetes was present before operation, American College of Anesthesiologists classification, exercise tolerance, breath-holding test classification, urine output, total length of hospital stay, length of intensive care unit stay, and preoperative human N-terminal brain natriuretic peptide concentration; the follow-up data includes whether serious complications occurred within 30 days after operation. This invention included 174 patients, with a median age of 70 years. There were 56 females (32.2%) and 118 males (67.8%). Of these, 114 patients (65.5%) did not experience serious postoperative complications within 30 days, while 60 patients (34.5%) experienced serious postoperative complications within 30 days. The average weight was 60.8 kg, and the average height was 168 cm. The surgical procedures included 18 gastrointestinal surgeries, 84 pancreatic surgeries, 38 urological surgeries, 4 orthopedic surgeries, and 30 hepatobiliary surgeries. The average operation time was 156 minutes, and the average anesthesia time was 224 minutes. 142 patients were smokers. ; 112 patients consumed alcohol; the average total intraoperative fluid infusion was 2560 ml; the average intraoperative blood transfusion was 320 ml; 85 patients had preoperative hypertension; 45 patients had preoperative diabetes; 87 patients were classified as AAS grade 2 or AAS grade 3; 35 patients had exercise tolerance grade 1, 106 patients had exercise tolerance grade 2, and 33 patients had exercise tolerance grade 3; 13 patients had breath-hold test grade 1, 76 patients had breath-hold test grade 2, and 85 patients had breath-hold test grade 3; the average urine output was [not specified]; the average total length of hospital stay was 15.3 days; the average length of stay in the intensive care unit was 3.2 days; the average preoperative human N-terminal brain natriuretic peptide concentration was 263.4 pg / L.

[0030] Step 2: Use univariate logistic regression to perform variable significance analysis on the aforementioned perioperative clinical data to screen out potential characteristic variables; the results of the variable significance analysis of the perioperative clinical data using univariate logistic regression are shown in Table 1.

[0031] Table 1 Results of univariate logistic regression analysis

[0032] Variable name Risk ratio (95% confidence interval) p-value Preoperative human N-terminal brain natriuretic peptide concentration 1.003（1.001-1.004） 0.001* age 1.12（1.030-1.230） 0.011* gender Female (for reference) - - male 1.059（0.344-3.258） 0.921 Did the patient have hypertension before surgery? 1.698（1.555-5.192） 0.033* Did the patient have diabetes before surgery? 2.558（1.808-8.101） 0.021* American Association of Anesthesiologists Classification 7.466（1.631-34.170） 0.010* height 1.015（0.956-1.079） 0.623 weight 1.068（1.009-1.130） 0.024* Breath-holding test classification 1.771（0.866-3.622） 0.092* Exercise endurance 0.585（0.347-1.389） 0.100* Duration of anesthesia 1.006（1.001-1.011） 0.022* Surgery duration 1.006（1.002-1.036） 0.360 Total intraoperative fluid volume 1.000（1.000-1.001） 0.044* Intraoperative blood transfusion volume 1.003（1.001-1.005） 0.009* Urine volume 1.854（0.714-3.662） 0.203 Total days of hospitalization 1.624（0.954-4.558） 0.084* Intensive Care Unit Days 1.525（0.881-4.568） 0.072* smoking history 2.324（0.547-5.280） 0.062* drinking history 3.122（0.358-7.844） 0.352 surgical site Liver surgery (reference) - - Pancreatic surgery 4.000（0.652-24.554） 0.134 Gastrointestinal surgery 1.680（0.342-8.259） 0.523 urological surgery - 0.998 Orthopedic surgery - 0.999

[0033] As shown in Table 1, the P-values ​​in Table 1 were obtained using univariate logistic regression analysis. According to statistical theory, if the P-value is ≤0.1, it indicates that the variable is a potential characteristic variable associated with severe postoperative complications in elderly patients. Based on the P-value results, perioperative clinical data including gender, age, weight, height, surgical site, operation duration, anesthesia duration, smoking history, alcohol consumption history, total intraoperative fluid volume, intraoperative blood transfusion volume, preoperative hypertension, preoperative diabetes, American College of Anesthesiologists (ACAS) classification, exercise tolerance, breath-holding test grade, urine output, total length of hospital stay, ICU length of hospital stay, and preoperative human N-terminal brain natriuretic peptide (NTP) concentration were screened. The potential characteristic variables identified were age, weight, anesthesia duration, smoking history, total intraoperative fluid volume, intraoperative blood transfusion volume, preoperative hypertension, preoperative diabetes, ACAS classification, exercise tolerance, breath-holding test grade, total length of hospital stay, ICU length of hospital stay, and preoperative human NTP concentration.

[0034] Step 3: Calculate the importance of the latent feature variables selected in Step 2 using optimal subset regression, and rank the latent feature variables selected in Step 2 according to their importance; in this embodiment of the invention, the latent feature variables are ranked according to their importance in the following order: preoperative human N-terminal brain natriuretic peptide concentration, age, American College of Anesthesiologists classification, duration of anesthesia, weight, intraoperative blood transfusion volume, preoperative hypertension, preoperative diabetes, exercise tolerance, total length of hospital stay, length of intensive care unit stay, smoking history, total intraoperative fluid infusion volume, and breath-holding test classification.

[0035] Step 4: Based on the reordered latent feature variables from Step 3, construct risk probability prediction models for different variable combinations using multifactor logistic regression. Model 1 includes only the first latent feature variable, Model 2 includes both the first and second latent feature variables, Model 3 includes the first, second, and third latent feature variables, and so on. In this embodiment of the invention, as the number of variables increases, the AIC value of the model gradually decreases. When the number of variables increases to six, the AIC value no longer decreases.

[0036] Step 5: Compare the risk probability prediction models of different variable combinations in Step 4 using the AIC value, and select the variable combination model with the smallest AIC value as the optimal risk probability prediction model. In this embodiment of the invention, the variable combination of the optimal risk probability prediction model is: preoperative N-terminal brain natriuretic peptide concentration, age, American College of Anesthesiologists classification, anesthesia duration, and weight.

[0038] Step Six: Evaluate the accuracy of the optimal risk probability prediction model described in Step Five using the ROC curve method. According to statistical theory, the area under the ROC curve is between 0.5 and 1.0. If the area under the ROC curve is greater than 0.7, the model's accuracy is considered clinically significant; if the area under the ROC curve is greater than 0.8, the model's accuracy is considered good and has high clinical significance. In this embodiment of the invention, the area under the ROC curve of the optimal risk probability prediction model is 0.899. Figure 3 As shown, the optimal risk probability prediction model provided by the embodiments of the present invention has good accuracy.

[0039] Step 7: Use the nomogram method to draw a nomogram model of the optimal risk probability prediction model described in Step 5. The nomogram is as follows: Figure 1 As shown.

[0040] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a nomogram model for predicting the probability of severe postoperative complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide, characterized in that, The steps are as follows: Step 1: Collect perioperative clinical and follow-up data of elderly patients; the perioperative clinical data of elderly patients includes gender, age, weight, height, surgical site, operation duration, anesthesia duration, smoking history, alcohol consumption history, types and dosages of intraoperative medications, total intraoperative fluid infusion, intraoperative blood transfusion, whether hypertension or diabetes was present preoperatively, American College of Anesthesiologists classification, exercise tolerance, breath-holding test classification, urine output, total length of hospital stay, length of intensive care unit stay, and preoperative human N-terminal brain natriuretic peptide concentration; the follow-up data includes whether serious complications occurred within 30 days postoperatively; Step 2: Use univariate logistic regression to perform variable significance analysis on the aforementioned perioperative clinical data to screen out potential characteristic variables; Step 3: Calculate the importance of the latent feature variables selected in Step 2 using optimal subset regression, and rank the latent feature variables selected in Step 2 according to their importance; Step 4: Based on the reordered latent feature variables from Step 3, construct risk probability prediction models for different variable combinations using the multi-factor logistic regression method. Model 1 includes only the first latent feature variable, Model 2 includes the first and second latent feature variables, Model 3 includes the first, second, and third latent feature variables, and so on. Calculate the AIC value of the risk probability prediction models for different variable combinations. Step 5: Compare the risk probability prediction models of different variable combinations in Step 4 using the AIC value, and take the variable combination model with the smallest AIC value as the optimal risk probability prediction model. Step Six: Evaluate the accuracy of the optimal risk probability prediction model described in Step Five using the ROC curve method; Step 7: Use the nomogram method to draw the nomogram model of the optimal risk probability prediction model described in Step 5.

2. The method for constructing a nomogram model for predicting the probability of postoperative serious complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide, as described in claim 1, is characterized in that... The potential characteristic variables selected include age, weight, duration of anesthesia, smoking history, total intraoperative fluid infusion, intraoperative blood transfusion, preoperative hypertension, preoperative diabetes, American College of Anesthesiologists classification, exercise tolerance, breath-holding test classification, total length of hospital stay, length of intensive care unit stay, and preoperative human N-terminal brain natriuretic peptide concentration.

3. The method for constructing a nomogram model for predicting the probability of postoperative serious complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide, as described in claim 1, is characterized in that... The potential characteristic variables selected were ranked according to their importance as follows: preoperative human N-terminal brain natriuretic peptide concentration, age, American College of Anesthesiologists classification, duration of anesthesia, weight, intraoperative blood transfusion volume, preoperative hypertension, preoperative diabetes, exercise tolerance, total length of hospital stay, length of intensive care unit stay, smoking history, total intraoperative fluid volume, and breath-holding test classification.

4. The method for constructing a nomogram model for predicting the probability of postoperative serious complications in elderly patients based on preoperative human N-terminal brain natriuretic peptide, as described in claim 1, is characterized in that... The optimal risk probability prediction model includes the following characteristic variables: preoperative N-terminal brain natriuretic peptide concentration, age, American College of Anesthesiologists classification, duration of anesthesia, and weight.

Images